View
0
Download
0
Category
Preview:
Citation preview
i
URBAN FLUX AND CONCENTRATION MEASUREMENTS
OF VOLATILE ORGANIC COMPOUNDS
AND CO2 IN MEXICO CITY
By
HECTOR ERIK VELASCO SALDAÑA
A dissertation submitted in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
WASHINGTON STATE UNIVERSITY Department of Civil and Environmental Engineering
DECEMBER 2005
ii
To the Faculty of Washington State University:
The members of the Committee appointed to examine the dissertation of HECTOR ERIK VELASCO SALDAÑA find it satisfactory and recommend that it be accepted.
___________________________________ Chair ___________________________________ ___________________________________ ___________________________________
iii
ACKNOWLEDGMENTS
First and foremost I would like to acknowledge my two primary advisors, Dr. Hal Westberg
and Dr. Brian Lamb for their guidance, encouragement and support throughout the entire process
of this dissertation. This dissertation represented to me a unique and magnificent opportunity to
develop my skills as engineer and researcher, meeting my professional objective to help to solve
the air pollution problem of Mexico City. The teaching and direction of my other advisors, Dr.
George Mount and Dr. Tom Jobson were fundamental for this research and my academic
formation, as well as all the lab and field support provided by Gene Allwine. Also, I would like
to express my appreciation to the rest of the Laboratory of Atmospheric Research (LAR), fellow
graduate students and friends, in particular to Shelley Pressley, Obie Cambaliza, Tara Strand and
Jack Chen, who made of my student life a wonderful experience.
A special acknowledgement to Dr. Luisa Molina and Dr. Mario Molina, leaders of the
MCMA-2003 field campaign and of the Integrated Program on Urban, Regional and Global Air
Pollution Program of the Massachusetts Institute of Technology (MIT), who have given me the
opportunity to collaborate in their project of cleaning the Mexican sky. Also, other special
acknowledgement is for the colleagues of the National Center for Environmental Research and
Training (CENICA) of the National Institute of Ecology of Mexico (INE) and for the colleagues
of the Atmospheric Monitoring Network of Mexico City (RAMA) for all their assistance in this
research.
In particular, I would like to thank the following colleagues that through their assistance,
discussion, data and expertise made possible this research: Miguel Zavala and Benjamin de Foy
(MIT), Berk Knigthon (Montana State University), Scott Herndon and Charles Kolb (Aerodyne),
Michael Alexander and Peter Prazeller (Pacific Northwest National Laboratory), Gustavo Sosa
(National Institute of Petroleum), Michael Grutter (National Autonomous University of Mexico),
William Brune (Pennsylvania State University), Victor Gutierrez-Avedoy, Adrian Fernandez,
iv
Felipe Angeles, Salvador Blanco and Rosa Maria Bernabe (CENICA & INE) and Victor Hugo
Paramo, Rafael Ramos and Armando Retama (RAMA).
An additional acknowledgement is for the National Commission of Science and Technology
of Mexico (CONACyT), which has followed and supported all my graduate formation. My
gratitude with CONACyT, which represents the scientific interests of the Mexican people, is a
responsibility that I will be happy to carry with me throughout my entire life.
Last but not least, I thank my family to make of me the person who I am. And specially, I
thank Elvagris Segovia for all her emotional support, demonstrating that the Pacific Ocean was
not a barrier for our love.
v
ATTRIBUTIONS
The research presented in this dissertation is the result of the participation of Washington
State University (WSU) in the MCMA-2002 & 2003 field campaigns in Mexico City. Both
campaigns were funded by the Integrated Program on Urban, Regional and Global Air Pollution
Program of the Massachusetts Institute of Technology (MIT) and the Metropolitan Commission
of Environment of Mexico City (CAM). Erik Velasco was partially funded by the National
Commission of Science and Technology of Mexico (CONACyT) with assistance for housing and
living.
The research contained in this dissertation is a compilation of many ideas and data collection
efforts from many people. Dr. Brian Lamb and Dr. Hal Westberg were invited to participate in
the MCMA-2002 and 2003 field campaigns and both prepared the initial proposal to measure
concentrations and fluxes of volatile organic compounds (VOCs) in the urban core of Mexico
City. In the course of the research, new goals and activities were added, including collaborations
with other groups involved in the MCMA field campaigns. Although multiple authors
contributed to the three manuscripts presented in this dissertation, Erik Velasco was the primary
author of all of them. Erik Velasco was responsible for essentially all of the data analysis and
had a primary role in many of the measurements performed during 2003.
Chapter 2 presents an integrating manuscript of several VOC measurements performed in the
2002 and 2003 field campaigns by different research groups at different locations. Data
contributions were made by Erik Velasco, Brian Lamb, Hal Westberg and Gene Allwine of
WSU, Tom Jobson, Michael Alexander and Peter Prazeller of the Pacific Northwest National
Laboratory (PNNL), Berk Knigthon of Montana State University (MSU), Scott Herndon and
Charles Kolb of Aerodyne, Jose Luis Arriaga and Gustavo Sosa of the National Institute of
Petroleum (IMP), Michael Grutter of the National Autonomous University of Mexico (UNAM)
and Miguel Zavala, Benjamin de Foy, Luisa Molina and Mario Molina of the Massachusetts
Institute of Technology (MIT). Chapter 3 is a paper that has been published and which presents
vi
results of the VOC flux measurements performed in 2003. This manuscript contains
contributions of Erik Velasco, Shelley Pressley, Brian Lamb, Hal Westberg and Gene Allwine of
WSU, Tom Jobson, Michael Alexander and Peter Prazeller of PNNL and Luisa Molina and
Mario Molina of the MIT. Chapter 4 is a paper that has been published and presents CO2 flux
measurements conducted by WSU in 2003. The authors of this paper include Erik Velasco,
Shelley Pressley, Gene Allwine Hal Westberg and Brian Lamb.
vii
URBAN FLUX AND CONCENTRATION MEASUREMENTS OF VOLATILE
ORGANIC COMPOUNDS AND CO2 IN MEXICO CITY
Abstract
by Hector Erik Velasco Saldaña, Ph.D.
Washington State University December 2005
Chair: Hal Westberg
Measurements of ambient concentrations and fluxes of volatile organic compounds (VOCs)
and CO2 in the atmosphere of Mexico City are reported here. These measurements were part of
the MCMA-2002 and 2003 field campaigns. In both campaigns, a wide array of VOC
measurements were conducted using a variety of methods with different spatial and temporal
scales at locations in the urban core, in a heavily industrial area and at boundary sites. The VOC
data were analyzed to document their distribution, diurnal pattern, origin and reactivity in
Mexico City. In the 2003 field campaign, an eddy covariance flux system was deployed at the
CENICA site to perform direct measurements of emissions of CO2 and selected VOCs (olefins,
acetone, methanol, toluene and C2-benzenes) from sources in an urban neighborhood. This work
demonstrates the use of micrometeorological techniques coupled with fast-response sensors to
measure fluxes of trace gases from urban landscapes, where the spatial variability of emission
sources, surface cover and roughness is very complex. The capability to evaluate emission
inventories using these techniques as described in this work is a valuable and new tool for
improving air quality management.
Overall, it was found that the VOC burden is dominated by alkanes (60%), followed by
aromatics (15%) and olefins (5%). However, in terms of ozone production olefins are the most
important hydrocarbons in Mexico City. Ambient concentrations and fluxes of VOCs and CO2
exhibited clear diurnal patterns with distinct correlation with vehicular traffic. The flux
viii
measurements showed that the urban landscape is nearly always a net source of VOCs and CO2.
The exception was acetone, which showed negative fluxes before sunrise. Fluxes of olefins,
acetone, toluene and C2-benzene were compared to the emissions reported in the local emissions
inventory. For the grid where the CENICA site was located, the emissions inventory generally
agreed with the measured fluxes. Furthermore, a comparison of the ambient VOC
concentrations in terms of lumped modeling VOC classes to the emissions inventory suggests
that some, but not all classes are underestimated in the inventory by factors of 3, as suggested in
previous studies.
ix
TABLE OF CONTENTS Page ACKNOWLEDGEMENTS................................................................................................ iii ATTRIBUTIONS .................................................................................................................v ABSTRACT........................................................................................................................vii LIST OF CONTENTS .........................................................................................................ix LIST OF TABLES............................................................................................................ xiii LISTOF FIGURES .............................................................................................................xv LISTOF ABBREVIATIONS..............................................................................................xii CHAPTER 1: INTRODUCTION
1.1. Overview...............................................................................................................1
1.2. Objectives .............................................................................................................4
1.3. Megacities and urban pollution.............................................................................5
1.3.1. Mexico City, a suitable megacity for air quality studies .............................6
1.3.2. Mexico City background..............................................................................7
1.4. Emission inventories.............................................................................................9
1.4.1. Application of micrometeorological techniques to evaluate atmospheric emissions.............................................................................................................11
1.5. VOCs: photochemical precursors and toxic pollutants.......................................12
1.6. CO2, a greenhouse gas ........................................................................................14
1.7. References...........................................................................................................15
CHAPTER 2: DISTRIBUTION, MAGNITUDES, REACTIVITIES, RATIOS AND DIURNAL PATTERNS OF VOLATILE ORGANIC COMPOUNDS IN THE VALLEY OF MEXICO DURING THE MCMA 2002 & 2003 FIELD CAMPAIGNS
2.1. Abstract ...............................................................................................................20
2.2. Introduction.........................................................................................................21
2.3. Monitoring sites ..................................................................................................24
2.4. Instrumentation ...................................................................................................25
x
2.4.1. Gas chromatography separation and flame ionization detection
(GC-FID)..............................................................................................................25
2.4.2. The Proton Transfer Reaction Mass Spectrometry (PTR-MS)..................26
2.4.3. Fast Olefin Sensor (FOS)...........................................................................27
2.4.4. Differential optical Absorption Spectroscopy (DOAS).............................28
2.5. Ambient air inter-comparison of GC, PTR-MS and DOAS
measurements for selected VOCs .......................................................................28
2.6. Results and discussion ........................................................................................30
2.6.1. Diurnal patterns of various VOCs at selected sites....................................30
2.6.2. Ambient mixing ratios and hydrocarbon reactivity ...................................32
2.6.3. Distribution of VOCs by groups ................................................................35
2.6.4. Comparison of ambient VOC concentrations to vehicles
exhaust signatures ................................................................................................37
2.6.5. Comparison of ambient VOC concentrations to the
emissions inventory .............................................................................................41
2.6.6. Comparison of olefin concentrations measured by FOS to olefin
concentrations calculated by the CIT model........................................................42
2.7. Summary and conclusions ..................................................................................44
2.8. References...........................................................................................................46
CHAPTER 3: FLUX MEASUREMENTS OF VOLATILE ORGANIC COMPOUNDS FROM AN URBAN LANDSCAPE
3.1. Abstract ...............................................................................................................70
3.2. Introduction.........................................................................................................70
3.3. Experimental methods ........................................................................................71
3.4. Results and discussion ........................................................................................73
3.5. Concluding remarks ............................................................................................76
3.6. Acknowledgements.............................................................................................77
3.7. References...........................................................................................................77
3.A. Auxiliary material ..............................................................................................83
3.A.1. Evaluation of emission inventories ...........................................................83
3.A.2. Micrometeorological techniques to measure fluxes of trace gases...........83
3.A.3. Field experiment .......................................................................................84
xi
3.A.3.1. The CENICA super site ................................................................84
3.A.3.2. The study period............................................................................85
3.A.3.3. Instrumentation .............................................................................85
3.A.3.4. The Fast Olefin Sensor..................................................................86
3.A.3.5. The Proton Transfer Reaction Mass Spectrometer .......................87
3.A.3.6. The open-path Infra Red Gas Analyzer (IRGA)...........................88
3.A.4. Eddy covariance flux technique................................................................89
3.A.5. Disjunct eddy covariance technique .........................................................90
3.A.6. Footprint analysis......................................................................................91
3.A.7. Validity for the eddy covariance system...................................................93
3.A.7.1. Statistical characteristics of the raw instantaneous
measurements..............................................................................................93
3.A.7.2. Spectral and cospectral analysis....................................................94
3.A.7.3. Stationarity test .............................................................................95
3.A.8. Evaluation of random and systematic errors in the measured
daily mean olefin flux .........................................................................................96
3.A.9. Diurnal profile of olefin concentration ....................................................98
3.A.10. Similarity between fluxes of VOC and CO2, and their
relationship with vehicular traffic.......................................................................98
3.A.11. Comparison between the measured VOC fluxes and VOC
emissions used for air quality models in Mexico City........................................99
3.A.12. References............................................................................................102
CHAPTER 4: MEASUREMENTS OF CO2 FLUXES FROM THE MEXICO CITY URBAN LANDSCAPE
4.1. Abstract .............................................................................................................121
4.2. Introduction.......................................................................................................121
4.3. Methods.............................................................................................................123
4.3.1. Measurements site and study period ........................................................123
4.3.2. Instrumentation ........................................................................................124
4.3.3. Postprocessing for eddy covariance flux calculations .............................125
4.3.4. Spectral and cospectra analysis................................................................127
4.3.5. Stationary test...........................................................................................128
xii
4.3.6. Footprint analysis.....................................................................................129
4.4. Results..............................................................................................................130
4.4.1. Concentrations .........................................................................................130
4.4.2. Fluxes.......................................................................................................131
4.4.2.1. Fluxes as a function of upwind direction.....................................133
4.4.2.2. Fluxes as a function of the vehicular activity ..............................133
4.4.2.3. Evaluation of random and systematic errors in the measured
daily mean CO2 flux .................................................................................134
4.5. Conclusions......................................................................................................136
4.6. Acknowledgements..........................................................................................137
4.7. References........................................................................................................138
CHAPTER 5: SUMMARY AND CONCLUSIONS
5.1. Summary ...........................................................................................................153
5.2. Conclusions.......................................................................................................157
5.3. Advantages and disadvantages of a flux system in an urban area ....................158
5.4. Future work.......................................................................................................159
xiii
LIST OF TABLES
CHAPTER 2
Table 2.1. Description of the VOC monitored sites during the MCMA-2002 and 2003
field campaigns ...................................................................................................................52
Table 2.2. Sensitivities, average concentrations measured during selected days
throughout the MCMA-2003 campaign between 6 and 10 am by a canister sampling
system and GC-FID analysis, FOS responses to those average concentrations, and
relative sensitivities to propylene for six compounds..........................................................53
Table 2.3. OH reactivity and average ambient concentrations of major VOCs measured
during the MCMA-2002 and 2003 studies at urban (Pedregal, La Merced, CENICA and
Constituyentes), rural (Santa Ana Tlacotenco, Tehotihuacan and La Reforma) and
industrial (Xalostoc) sites of the Valley of Mexico. Data correspond to the morning
rush hours (6 to 9:00 h)........................................................................................................54
Table 2.4. Ambient average concentrations of halogenated VOCs measured during the
MCMA-2003 study at urban (Pedregal, La Merced and CENICA) and industrial
(Xalostoc) sites of the Valley of Mexico. ............................................................................56
Table 2.5. Hydrocarbon molar ratios (ppbC/ppbC) measured at the industrial, urban and
rural sites and from vehicle chase operation during the MCMA-2002 and 2003 field
studies. Also ratios from vehicles exhaust studies in North America are shown for
comparisons ........................................................................................................................57
Table 2.6. Comparison between ambient VOC concentrations measured at urban sites
during the morning period (6 to 9:00 h) and the corresponding VOC emissions from the
most recent emissions inventory for modeling purposes ....................................................58
xiv
CHAPTER 3
Table 3.A.1. Sensitivities, average concentrations measured during selected days
throughout the campaign between 6 and 10 am from the sampling line of the eddy
covariance system by a canister sampling system, FOS responses to those average
concentrations, and relative sensitivities to propylene for the 6 evaluated species ..........106
Table 3.A.2. Average ratios of the normalized fluxes of each analyzed VOC and
CO2 for different periods of the diurnal cycle. A high ratio indicates that the VOC
specie and CO2 are emitted by the same sources, while a low ratio indicates that the
VOC and CO2 species are emitted by different sources ...................................................107
xv
LIST OF FIGURES
CHAPTER 1
Figure 1.1. Map of the Valley of Mexico showing the Mexico City Metropolitan
Area and associated information......................................................................................................7
CHAPTER 2
Figure 2.1. Sampling sites during MCMA-2002 and MCMA-2003 field campaigns. Points
indicate the location of the sampling sites and numbers correspond to the sites listed in Table 1.
The gray scale shows the orography of the Valley of Mexico and the black contour limits the
Federal District...............................................................................................................................59
Figure 2.2. Time series of benzene (a) and toluene (b) mixing ratios measured by DOAS, PTR-
MS and GC-FID. The resolution time for DOAS was 5 minutes and for PTR-MS ~30 sec.
PTR-MS points correspond to 1 minute averages. Samples collected by canisters and analyzed
by GC-FID correspond to 1 hour averaging. Short term spikes were commonly observed with
PTR-MS indicating local sources ..................................................................................................60
Figure 2.3. Time series of C2-benzenes and C3-benzenes measured by the PTR-MS onboard a
mobile laboratory together with GC-FID samples collected in parallel at La Merced (a, b),
Pedregal (c, d) and Santa Ana sites (e, f). PTR-MS concentrations represent hourly averages.
GC-FID samples were collected in hourly periods at La Merced and Pedregal sites, and in 3-
hours periods at Santa Ana Tlacotenco. The dashed lines indicate ±1 standard deviation of the
PTR-MS concentrations.................................................................................................................61
Figure 2.4. Average diurnal pattern of the olefinic mixing ratio (as propylene) detected by the
FOS for 23 days during the MCMA-2003 study (black line) and for individual weeks. The gray
shadow represents the one standard deviation range, and gives and indication of the day-to-day
variability in each phase of the daily cycle....................................................................................62
xvi
Figure 2.5. VOC distribution by groups during the morning (a) and afternoon (b). Numbers in
the columns indicate the percent contribution of each VOC group to the total VOC
concentration, which is displayed at the bottom of each column in ppbC ....................................63
Figure 2.6. Correlations between alkenes comparing ambient data to vehicle chase samples.
Ambient data correspond to all canister samples and the dashed line indicates the regression line
for vehicular emissions ..................................................................................................................64
Figure 2.7. Correlations between alkanes comparing ambient data to vehicle chase samples.
Ambient data correspond to all canister samples and the dashed line indicates the regression line
for vehicular emissions ..................................................................................................................65
Figure 2.8. Correlations between aromatics comparing ambient data to vehicle chase samples.
Ambient data correspond to all canister samples and the dashed line indicates the regression line
for vehicular emissions ..................................................................................................................66
Figure 2.9. Comparison of urban and vehicle exhaust hydrocarbon abundances relative to
acetylene in (ppbC/ppbC). The closed circles indicate the median values for vehicle exhaust,
while the open circles indicate the median values for urban data collected from urban sites
(Pedregal, La Merced and CENICA) between 6 and 9:00 h. The gray shading encloses the 10th
and 90th percentiles of the urban values.........................................................................................67
Figure 2.10. Olefins concentration measured by the FOS and calculated by the CIT model (a)
during 5 days and (b) average concentrations measured during the entire MCMA-2003 field
campaign at the CENICA site. The gray shading indicates the ±1 standard deviation range from
the FOS measurements ..................................................................................................................68
CHAPTER 3
Figure 3.1. Left side: emissions from different sources are blended by the turbulence
generated in the roughness sub-layer, producing a net exchange flux in the constant flux
layer where the flux is measured. Right side: fraction of the flux measured (F/S0) at the
xvii
tower height versus the upwind distance or effective source footprint (x). F is the flux
and S0 the source strength .............................................................................................................79
Figure 3.2. Average diurnal pattern of olefin fluxes predicted and measured for the 23-
day study, and for weekdays and weekends. The dashed line indicates the sum of
olefinic VOC emissions reported in the emissions inventory for the grid cell where the
flux system was located. The grey shadow represents ±1 standard deviation from the
total average ..................................................................................................................................80
Figure 3.3. Olefin flux distribution as a function of the upwind direction during the
entire study. The contour indicates the footprint that is estimated to encompass 80% of
the total flux. Flux intensity is shown in grey scale .....................................................................81
Figure 3.4. Average diurnal patterns of the flux of (A) methanol, (B) acetone, (C)
toluene and (D) C2-benzenes measured by the PTR-MS during ten days of the study.
The grey shadows represent ±1 standard deviation from the total averages ................................82
Figure 3.A.1. Schematic diagram of the instrumented flux tower. The azimuth
orientation of the sonic anemometer is 16º from north. Dimensions are in meters ....................108
Figure 3.A.2. Comparison of identified olefins weighted by their corresponding FOS
sensitivities from GC-FID measurements versus FOS measurements for 21 sample
periods during selected days throughout the campaign between 6 and 10 am ...........................109
Figure 3.A.3. Comparison between olefin fluxes (µg m-2 s-1) calculated using the
techniques of eddy covariance and disjunct eddy covariance with different sampling
intervals (a) 0.6 s, (b) 1.2 s, (c) 2.4 s and (d) 3.6 s ......................................................................110
Figure 3.A.4. Different fractions of the measured flux (F/S0) during the entire campaign
as function of the wind direction for different intervals of time, (a) from 0:00 to 6:00 h,
(b) from 6:00 to 9:00 h, (c) from 9:00 to 15:00 h and (d) from 14:00 to 21:00 h. ......................111
xviii
Figure 3.A.5. A 20-s trace of methanol and olefins mixing ratios plotted together with
vertical wind speed beginning at 9:40 h local time on April 11, 2003. ......................................112
Figure 3.A.6. Spectra for olefins mixing ratio (a) and cospectra for olefins mixing ratio
and vertical wind speed (b) for 6 different periods of 30 minutes before the low-pass
filter correction. The -5/3 and -7/3 slopes indicate the theoretical slopes of the inertial
subrange. ................................................................................................................................113
Figure 3.A.7. Stationarity test for olefin flux: a measured period is stationary if the
average flux from 6 continuous subperiods of 5 min is within 60% of the flux obtained
from a 30 min average. In our study 82% of the periods fulfilled this criterion. ......................114
Figure 3.A.8. Effects of random and systematic errors for the mean daily olefin flux. (a)
Effects due to random errors for various percentages of error. (b) Effects due to
systematic errors as a function of the error percentage. In both figures the horizontal line
represents the mean daily flux, 0.355 µg m-2 s-1. ........................................................................115
Figure 3.A.9. Average diurnal pattern of olefinic concentration detected by the FOS for
the entire study (black line) and for individual weeks. The gray shadow represents ±1
standard deviation from the total average, and gives an indication of the day-to-day
variability in each phase of the daily cycle. ................................................................................116
Figure 3.A.10. Diurnal profile of olefin fluxes and traffic counts for two intersections
within the tower footprint (intersection between Av. Ermita Iztapalapa and Av. Rojo
Gomez I1-2, and the intersection between Av. Ermita Iztapalapa and Av. Anillo
Periferico I2-3). .............................................................................................................................117
Figure 3.A.11. Ratios between the normalized fluxes of VOC and CO2. (a) for olefins,
(b) methanol, (c) acetone, (d) toluene, and (e) C2-benzenes. ......................................................118
Figure 3.A.12. Comparison of the diurnal profiles of fluxes of toluene (a), C2-benzenes
(b), and acetone (c) with the diurnal profiles of emissions used in the CIT three-
xix
dimensional photochemical model for Mexico City. The dashed lines indicate ±1
standard deviation of the measured fluxes. .................................................................................119
CHAPTER 4
Figure 4.1. Schematic diagram of the instrumented flux tower. The azimuth orientation
of the sonic anemometer is 16º from north. Dimensions are in meters ......................................141
Figure 4.2. (a) Power density spectra for CO2 concentration and ambient temperature,
normalized for comparison. The -5/3 slope indicates the theoretical slope in the inertial
subrange. (b) Cospectra of vertical velocity with ambient temperature and CO2
concentration, normalized for comparison. X represents the average CO2 concentration
and ambient temperature for the spectra, and the covariances of those scalars with the
vertical wind speed for the cospectra. Overall, the shape and details of the CO2 and
ambient temperature spectra and cospectra correspond closely over the entire range of
frequencies measured. The data correspond to the 30 minute sampling period on April 7
at 15:30 h…..................................................................................................................................142
Figure 4.3. Stationarity test for CO2 flux: in 56% of the periods, the flux difference was
less than 30%, which indicates periods that meet and exceed the stationarity criteria. In
18% of the periods, the flux difference was between 30 and 60%, which means that
these periods have an acceptable quality .....................................................................................143
Figure 4.4. Fraction of the flux measured (F/S0) versus the upwind distance or effective
fetch (x)….… ...............................................................................................................................144
Figure 4.5. Different fractions of the measured flux (F/S0) during the entire campaign as
function of the wind direction for different intervals of time, (a) from 0:00 to 3:00 h, (b)
from 6:00 to 9:00 h, (c) from 12:00 to 15:00 and (d) from 18:00 to 21:00 h ..............................145
xx
Figure 4.6. Average diurnal pattern of CO2 concentrations for the entire study and for
separate weeks. The gray shadow represents ±1 standard deviation from the total
average…………………………………………………………………………………… ........146
Figure 4.7. Average diurnal pattern of CO2 fluxes for the entire study and for weekdays
and weekends. The gray shadow represents ±1 standard deviation from the total
average…….. ...............................................................................................................................147
Figure 4.8. CO2 flux distribution as a function of the upwind direction during the entire
study. The contour shape indicates the fraction of the flux measured equal to 80% and
the shading indicates the magnitude of the CO2 fluxes. The black spot indicates the
position of the flux tower; and the solid area in the left bottom corner represents part of
the National park “Cerro de la Estrella”. The four primary roads surrounding the
measurement site are: (1) Av. Rojo Gomez, (2) Av. Ermita Iztapalapa, (3) Anillo
Periferico, (4) Av. Jalisco, and (5) Calz. San Lorenzo. ..............................................................148
Figure 4.9. Diurnal profile of CO2 fluxes superimposed over plots of traffic counts for
two intersections within the footprint: intersection I1-2 between Av. Ermita Iztapalapa
and Av. Rojo Gomez, and intersection I2-3 between Av. Ermita Iztapalapa and Anillo
Periferico. Fluxes correspond to the upwind sectors where both intersections are
located, for I1-2 the sector between 240 and 285º and for I2-3 the sector between 115 and
160º ….…………….. ..................................................................................................................149
Figure 4.10. Correlation between CO2 fluxes and vehicular traffic for two intersections,
I1-2 and I2-3. Fluxes correspond to the 45º upwind sectors where both intersections are
located, respectively ....................................................................................................................150
Figure 4.11. Effects of random and systematic errors for the mean daily CO2 flux. (a)
Effects due to random errors for various percentages of error. (b) Effects due to
xxi
systematic errors as a function of the error percentage (ps). In both figures the
horizontal line represents the mean daily CO2 flux, 0.41 mg m-2 s-1.. .........................................151
xxii
LIST OF ABBREVIATIONS
ATDD Atmospheric Turbulence and Diffusion Division of NOAA BTEX sum of benzene, toluene, ethylbenzene and xylenes BTX sum of benzene, toluene and xylenes BVOC biogenic VOC CAM Metropolitan Environmental Commission of Mexico City CENICA National Center for Environmental Research and Training CIT California Institute of Technology airshed model cχ’ instantaneous deviation of the trace gas d zero displacement plane DEA disjunct eddy accumulation DEC disjunct eddy covariance DOAS UV Differential Optical Absorption Spectrometer EC eddy covariance EPA Environmental Protection Agency ETBE Ethyl tertiary-butyl ether FFT Fast Fourier Transform FID flame ionization detector FOS Fast Olefin Sensor FTIR Fourier Transform Infrared Spectrometer Fχ flux of a trace gas GC gas chromatography GC-FID gas chromatography/flame ionization detection GC-MS gas chromatography/mass spectroscopy detection I1-2 intersections between avenues 1 and 2 in Figure 3.3 I2-3 intersections between avenues 2 and 3 in Figure 3.3 IMP Mexican Petroleum Institute INE Mexican National Institute of Ecology IRGA open-path InfraRed Gas Analyzer k Von Karman constant L Monin-Obuhkov length LPG Liquefied petroleum gas MCMA Metropolitan Area of Mexico City MCMA-2002 Metropolitan Area of Mexico City 2002 field campaign MCMA-2003 Metropolitan Area of Mexico City 2003 field campaign MILAGRO Megacity Initiative: Local and Global Research Observations MIT Massachusetts Institute of Technology MS mass spectroscopy MSU Montana State University MTBE Methyl tertiary-butyl ether NOAA National Oceanic and Atmospheric Administration PM particulate matter PM10 particulate matter with an aerodynamic diameter equal or smaller to 10 µm PM2.5 particulate matter with an aerodynamic diameter equal or smaller to 2.5 µm
xxiii
PNNL Pacific Northwest National Laboratory ppbv part per billion volume ppmv part per million volume pr percentage due to random error ps percentage due to systematic error PTR-MS Proton Transfer Reaction-Mass Spectrometer RAMA Automated Atmospheric Monitoring Network of Mexico City REA relaxed eddy accumulation S0 source strength T ambient temperature measured by the sonic anemometer TDLAS Tunable Diode Laser Spectroscopy TOC total organic compounds u mean wind velocity UNAM National Autonomous University of Mexico VOCs volatile organic compounds w vertical wind speed WSU Washington State University w’ instantaneous deviation of the vertical wind velocity x upwind distance or effective source footprint z0 surface roughness parameter zcity height of the urban canopy zm measurement height ztower height of the flux tower ξ atmospheric stability index
xxiv
1
CHAPTER 1
INTRODUCTION
1.1. Overview
Poor air quality in an increasingly urbanized world directly threatens the health of a large
fraction of the world’s population. Increases in pollutant concentrations impact the viability of
important natural and agricultural ecosystems in regions surrounding highly urbanized areas, and
increasing emissions of radiatively important trace gases, such as carbon dioxide (CO2),
contribute significantly to global climate change. These issues are particularly acute in the
developing world where the rapid growth of megacities is producing atmospheric pollution of
unprecedented severity and extent. It has been recognized that in developing megacities the
main sources of pollutants are vehicles and industries. However, there are a number of other
significant sources that cannot be identified or quantified accurately. These include unregulated
factories and domestic activities, such as residential cooking and heating. Uncertainties in
emissions are particularly large in third world megacities for volatile organic compounds (VOCs)
and CO2. The VOCs are reactive species that contribute to the creation of photochemical smog
and tropospheric ozone (O3), while CO2 is the most significant anthropogenic greenhouse gas.
With the aim of improving our understanding of the emissions of VOCs and CO2 in megacities,
we participated in the Mexico City Metropolitan Area (MCMA) field campaigns in 2002 and
2003. Our participation involved two main activities. First we investigated the distribution and
diurnal pattern of VOCs at several different sites within and near Mexico City. VOC signatures
for airshed boundary sites, central urban core and downwind urban receptor sites were obtained
from ambient samples collected using canisters and analyzed by gas chromatography/flame
ionization detection (GC-FID) methods. Continuous olefin mixing ratios were measured from a
tower erected on a laboratory building in southeastern Mexico City (CENICA super site) using a
fast chemiluminsecent olefin sensor (FOS). Additional diurnal VOC concentration
measurements made by other participants included use of a Proton Transfer Reaction-Mass
2
Spectrometer (PTR-MS) onboard a mobile laboratory, a second PTR-MS located at the CENICA
super site, and long path measurements of selected VOCs using UV Differential Optical
Absorption Spectrometers (DOAS) and Fourier Transform Infrared Spectrometers (FTIR)
located near of the center of the city and at the CENICA super site.
Our second activity involved the deployment of micrometeorological flux instruments on the
tower at the CENICA supersite to directly measure fluxes of VOCs and CO2 from the urban
landscape. Fluxes of olefinic VOCs were measured using the FOS and processed by the eddy
covariance technique (EC). Fluxes of acetone, methanol, C2-benzenes and toluene were
measured using the PTR-MS and computed by the disjunct eddy covariance technique (DEC).
Fluxes of CO2 and water vapor were measured by an open path infrared sensor and processed by
EC. The stationarity and frequency distribution of turbulence for the flux measurements were
examined to demonstrate the applicability of micrometeorological techniques to the
heterogenous urban landscape where the spatial variability of emission sources, surface cover,
and roughness is high. The measured fluxes were analyzed as functions of the upwind direction
and extent of the source footprint to identify the strongest sources. Traffic counts were used to
examine the relationship between fluxes of VOCs and CO2 with vehicular activity. Finally, the
measured fluxes were compared to emissions inventory data for the grid where the flux tower
was located, and the VOC data from the canisters samples were examined in terms of the relative
composition of lumped modeling VOC classes and compared to the composition from the most
recent emissions inventory used for modeling. At the end of this research, we expect to have
improved our ability to evaluate emission inventories of trace gases through direct and indirect
methods and to have demonstrated the feasibility of EC and DEC techniques to perform VOC
flux measurements in an urban area using state of the art VOC sensors. It is important to
recognize that this experiment was the first time that micrometeorological techniques were
applied to measure fluxes of VOCs in an urban landscape as a new tool for improving air quality
management.
3
The dissertation is divided into five chapters. The Introduction chapter provides a brief
overview of the work plan, lists the objectives and a short review of the importance of using
ambient concentrations to evaluate emissions inventories of VOCs and CO2 in urban
environments. The second chapter entitled “Distribution, magnitudes, reactivity, ratios and
diurnal patterns of volatile organic compounds in the Valley of Mexico during the MCMA 2002
& 2003 field campaigns” presents a summary of all ambient VOC measurements performed in
both MCMA field campaigns, analysis of the VOC data in terms of reactivity to form ozone,
VOC distributions and diurnal patterns, an examination and comparison of the ambient VOC
data with vehicle exhaust profiles, and a comparison in terms of lumped modeling VOC to the
actual emissions inventory. This chapter will be reformed and submitted to the Journal of
Atmospheric Chemistry and Physics for publication. The third and fourth chapters are
manuscripts that were published in specialized journals. The versions included here correspond
to the final drafts. The manuscript corresponding to the third chapter is entitled “Flux
measurements of volatile organic compounds from an urban landscape”. This manuscript
presents the flux measurements performed at the CENICA supersite as part of the MCMA-2003
field campaign. This manuscript was published in Geophysical Research Letters (Velasco et al.,
2005a). The fourth chapter is a manuscript entitled “Measurements of CO2 fluxes from the
Mexico City urban landscape”. The CO2 fluxes discussed in this manuscript were measured in
parallel with the VOC fluxes presented in chapter three. This manuscript was published in
Atmospheric Environment (Velasco et al., 2005b). Finally, the fifth chapter contains a brief
summary and presents the final remarks for this dissertation.
4
1.2. Objectives
The overall goal of this research was to improve our understanding of atmospheric emissions
in megacities and, thus, to improve our ability to evaluate emission inventories for a better
design of strategies to reduce air pollution. This study focused on VOC and CO2 emissions in
Mexico City. In particular, VOCs are important species in the evolution of ozone and other toxic
pollutants, while CO2 is known to be the most important greenhouse gas of anthropogenic origin.
In pursuing this goal, we addressed the objectives listed below.
1) To explore the use of micrometeorological techniques to evaluate emissions of all
anthropogenic and biogenic sources in urban landscapes.
2) To demonstrate the feasibility of using micrometeorological techniques to measure urban
fluxes of trace gases in urban environments where the spatial variability of surface cover,
emission sources and roughness is high.
3) To evaluate the use of state of the art fast-response sensors to measure VOC fluxes from
urban landscapes.
4) To determine the diurnal pattern of fluxes of CO2 and VOCs from a typical neighborhood of
Mexico City, and to determine their magnitudes as a function of the upwind direction and
footprint to identify the strongest sources.
5) To provide direct measurements of emissions for a number of VOCs for comparison to
existing gridded emission inventories.
6) To analyze the relationship between fluxes of both, VOCs and CO2 with the vehicular
activity.
7) To investigate the VOC signature and its diurnal pattern at selected urban sites in Mexico
City to improve our understanding of atmospheric chemistry and anthropogenic emissions.
8) To evaluate the underestimation of VOC emissions in the gridded emission inventory as
suggested in previous studies.
5
1.3. Megacities and air pollution
About half of the world’s population now lives in urban areas because of the opportunity for
a better quality of life. Many of these urban centers are expanding rapidly, leading to the growth
of megacities, which are defined as metropolitan areas with populations exceeding 10 million
inhabitants (Molina and Molina, 2004). At present, there are 20 megacities with a combined
population of 292 million (UN, 2004), and more than 100 smaller cities worldwide that have
similar characteristics. These agglomerations of people are leading to increased stress on the
natural environment, with impacts at urban, regional and global scales (Decker et al., 2000).
In recent decades, air pollution has become one of the most important problems of
megacities. In some cases, megacities emit plumes of pollutants at rates even greater in size than
entire countries (Gurjar and Lelieveld, 2005). These plumes contain large amounts of toxic
pollutants, including ozone precursors, aerosols and greenhouse gases, which produce serious
impacts on public health, ecosystems and global change. In different megacities around the
world, the adverse effects of pollution on human health have been demonstrated with similar
conclusions: particulate matter (PM), O3, and other air pollutants attack the cardiovascular and
respiratory systems and are associated with premature mortality as well as sickness (Dockery et
al., 1993; Evans et al., 2002). The export of air pollutants from a megacity to regional and global
environments is a major concern because of wide-ranging potential consequences for natural
ecosystems, visibility degradation, weather modification, changes in precipitation chemistry,
changes in radiative forcing and tropospheric oxidation or self-cleaning capacity (Jacob et al.,
1999; Chameides et al., 1994; Kiehl, 1999; Gauss et al., 2003; Prinn et al., 2001).
Clearly air pollution in megacities is an important issue. However, emission estimates for
megacities are highly uncertain and not much is known about atmospheric chemistry in the
tropics, where much of the urban world population growth occurs. This represents an enormous
scientific challenge due to the complexity of social, physical and chemical processes that must be
understood.
6
1.3.1. Mexico City, a suitable megacity for air quality studies
Although, no one megacity can fully represent all of world’s megacities, Mexico City is a
good location for investigations to improve our understanding of the atmospheric pollution in
megacities. Mexico City has the following properties:
Mexico City is the world’s second largest megacity after Tokyo, and will continue to be one
of the most populous cities, with a current population near 19 million.
Mexico City is situated in the tropics as are most of the world’s fastest growing megacities.
Its emission characteristics are roughly intermediate between those of a city in emerging
economies and those of a city from a developed country.
Extensive air quality measurements have been made for many years. Criteria pollutants have
been routinely monitored for more than 15 years. A number of intensive studies have been
performed during the last decade to provide insight into the local air quality problem.
Emissions inventories have been developed and are being refined.
Compared with other tropical cities, Mexico offers good infrastructure and logistics to
support the investigation of air quality in a megacity.
For the purposes of evaluating urban emissions of VOCs and CO2 through direct
measurements, Mexico City represents the ideal site for the reasons described above and because
its VOC and CO2 concentrations are among the highest in the world (Arriaga et al., 2004).
Emissions of these species from combustion sources are accentuated by the high elevation of the
city (2240 m.s.l.), which makes combustion processes less efficient. VOC evaporative emissions
from a variety of sources such as cleaning, painting, and industrial processes are large due to the
subtropical weather with temperatures above 20ºC and intense solar radiation year round. Lastly,
the main contributor to high VOC and CO2 concentrations in Mexico City is the extensive
presence of aged industrial operations and a relatively old vehicle fleet. Figure 1.1 shows a map
of the Valley of Mexico, where the metropolitan area and topography are shown.
7
Position: 19º25’ N latitude’ and 99º10’ W longitude
Population: ~ 19 million (2003)(a). 17.9% of the total Mexican population
Growth projection to 20.6 million (2015).
Second largest city in the world.
Density population: 12000 inhabitants per km2 (b)
Urbanized area: 1500 km2 (1999)(b)
(a) (UN, 2004), (b) (Molina et al., 2002).
GDP per capita: $7750 USA dollars (2000)(b)
Figure 1.1. Map of the Valley of Mexico showing the Mexico City Metropolitan Area and
associated information.
1.3.2. Mexico City background
Mexico City lies in an elevated basin at an elevation of 2240 m. The nearly flat basin covers
~5000 km2 of the Mexican Plateau and is confined on three sides (east, south, and west) by
mountain ridges, with a broad opening to the north and a narrower gap to the south-southwest.
The surrounding ridges vary in elevation, with several peaks reaching nearly 4000 m. Two
major volcanoes, Popocateptl (5452 m) and Ixtacihuatl (5284 m), are on the mountain ridge
southeast of the basin. The metropolitan area is on the southwest side of the basin and covers
~1500 km2.
Mexico City is also named the Mexico City Metropolitan Area (MCMA), since it combines a
conurbation proper with peripheral zones. The MCMA consists of 16 delegations of the Federal
District and 37 contiguous municipalities from the State of Mexico and one municipality from
8
the State of Hidalgo, some with populations over 1 million that make up the total population of ~
19 million.
The large population, 53,000 industries, 3.5 million vehicles, complex topography, and
meteorology cause high pollution levels. The mountains, together with frequent thermal
inversions, trap pollutants within the basin. The high elevation and intense sunlight also
contribute to photochemical processes that create O3 and other secondary pollutants. More than
47 million L of fuel consumed per day produce thousands of tons of pollutants. Air pollution is
generally worst in the winter, when rain is less common and inversions more frequent (Molina et
al., 2002).
Owing to the high altitude, MCMA air contains ~23% less oxygen than at sea level.
Consequently, internal combustion engines need to be carefully tuned to the proper O2-to-fuel
ratio to improve combustion efficiency and reduce emissions. People at higher altitudes are
more susceptible to respiratory ailments than those at sea level. More air must be inhaled for an
equivalent amount of O2 at high altitudes, which causes a higher dose of air pollutants (Molina
and Molina, 2004).
High O3 concentrations are measured throughout the year because the subtropical latitude
and high altitude are conducive to photochemistry. The national air quality standard for ozone
(110 ppb/1 hour average) is exceeded 70% of the days (GDF, 2004). Anticyclone high pressure
systems appear during winter, resulting in light winds above the basin and nearly cloudless skies.
This leads to the formation of strong surface-based inversions at night that persist for several
hours after sunrise. Strong solar heating of the ground generates turbulent mixing that erodes
these inversions in the morning, producing deep boundary layers by the afternoon. Pollutants
trapped below the inversion layer are then mixed within the convective boundary layer, which
can reach altitudes of 4 km. There is sufficient time for O3 formation in the morning before the
development of the deep convective boundary layer because of high emission rates and intense
solar radiation. During the wet summer months (June to September), clouds inhibit
photochemistry and rainfall removes many trace gases and PM; high O3 episodes are less
9
frequent. Near-surface northerly winds during the day may transport pollutants to the southwest,
where O3 concentrations are the highest (Bossert, 1997).
1.4. Emission inventories
The design of emission controls requires detailed information on the status of the air quality
(provided by monitoring networks) and the principal sources of pollution and their location, as
characterized by emission inventories. Combining the information from monitoring and
emission estimates with knowledge of dispersion characteristics for the city and chemical
transformations of pollutants enables air quality models to be developed. Such models are
powerful tools for air quality managers, but their results depend on the quality of the emissions
estimates and meteorological inputs. A comprehensive emissions inventory must provide an
accurate tally of all important air pollutant species and should include the temporal and spatial
distribution of emissions in a region.
Emission inventories are typically constructed through an aggregation process that accounts
for emission rates, activity levels, and the distribution of sources. Emissions inventories for
Mexico City have been under development since 1986 and are periodically updated by the
Mexican government. The most recent inventory, for the year 2002 by the Secretaria del Medio
Ambiente del Gobierno del Distrito Federal (SMA-GDF, 2005), was constructed bottom-up from
point, area, mobile and natural sources. Totals in Gg yr-1 are as follows: PM with an
aerodynamic diameter equal or smaller to 10 µm (PM10) ~ 23, PM with an aerodynamic diameter
equal or smaller to 2.5 µm (PM2.5) ~ 7, sulfur dioxide (SO2) ~ 9, carbon monoxide (CO) ~ 1941,
nitrogen oxides (NOx) ~ 186, total organic compounds (TOC) ~ 709, methane (CH4) ~ 164,
VOCs ~ 490, and ammonia (NH3) ~ 17. Substantial uncertainties exist, particularly concerning
the VOC emissions.
Emission rates are often derived from laboratory or specific field measurements (i.e. vehicle
dynamometer studies), activity levels can be obtained from traffic counts, surveys of sources and
other information, and source distributions may come from roadway maps, aerial photographs, or
10
be derived as a function of population density. The propagation of errors associated with this
bottom up inventory compilation can result in large uncertainties in urban areas, where it is
almost impossible to identify and quantify all sources. This is particularly true for VOC
emissions (Arriaga et al., 2004; Molina et al., 2002; Fujita et al., 1992). It has been recently
reported that the VOC emissions inventory in Mexico City may be underestimated by a factor of
3 based upon comparison of photochemical model results and observations (West et al., 2004).
Emissions inventories have been evaluated through ambient measurements. The most direct
evaluation has been to compare measured concentrations of NOx and VOC with model results
(West et al., 2004; Sillman, 1999; Trainer et al., 2000). Using this approach, input emissions in
models are modified to obtain the closest possible agreement with ambient measurements, with
the assumption being that all errors come from the emissions inventory and not from model or
measurements uncertainties. Other evaluation methods are based on measured ratios between
species (Goldan et al., 1995). For example, the morning VOC/NOx ratio has been demonstrated
to be useful for this purpose because chemical losses are relatively small during the morning and
ambient concentrations reflect the ratio of directly emitted species (Arriaga et al., 2004; Sillman,
1999). In addition, the chemical mass balance receptor modeling technique has been
successfully applied to improve emissions inventories by identifying the relative contribution of
different sources to measured VOC fingerprint (Watson et al., 2001). For the case of mobile
sources, emissions estimates have been evaluated using open path sensors positioned across
roads and from tunnel studies. In general, air quality measurements in tunnels have proven to be
useful to measure emissions from vehicles operating under real-world conditions (Sawyer et al.,
2000). VOCs and other species are not subject to photochemical degradation in the tunnel and
speciation data can be used to estimate the composition of vehicular emissions. Emission rates
derived from data collected during remote sensing and tunnel studies are based on the
relationship with CO2 from the same source. CO2 is the primary carbon-containing product of
fuel combustion and provides a measure of the amount of fuel burned. The concentrations of
emitted pollutants relative to the CO2 concentration provide measurements of these emissions per
11
quantity of fuel consumed (McGaughey et al., 2004). These fuel-based emission factors,
expressed as mass emitted per unit volume of fuel burned, are expected to vary less over the full
range of driving conditions than travel-based emission factors (Singer and Harley, 1996).
1.4.1. Application of micrometeorological techniques to evaluate atmospheric emissions
In the case of biogenic hydrocarbon emissions, micrometeorological techniques have been
successfully used to measure surface fluxes from woodlands and agricultural fields. Biogenic
VOC (BVOC) fluxes have been measured by techniques such as surface layer gradient, relaxed
eddy accumulation (REA), EC and DEC (Guenther et al., 1996; Westberg et al., 2001; Karl et
al., 2002). The first two are indirect measurements of flux that rely on empirical
parameterizations. EC is a direct flux measurement technique that requires a sampling rate of
approximately 5 Hz. Fast-response sensors have recently become available so that EC methods
can be used to measure VOC fluxes above a forest canopy (Guenther and Hills, 1998; Pressley,
2005) and over crop lands (Shaw et al., 1998). The capabilities of the PTR-MS have been
expanded to allow direct flux measurements of a number of BVOC through DEC (Spirig et al.,
2005; Karl et al., 2002; Rinne et al., 2001). This method calculates fluxes using measurements at
a lower frequency compared to EC; the sample interval can be as long as 30 s. This makes it
possible to use relatively slow sensors for flux measurements.
In urban areas, micrometeorological approaches have been previously applied to measure
fluxes of momentum, heat and CO2 (Grimmond et al., 2004; Moriwaki and Kanda et al., 2004;
Soegaard and Møller-Jensen, 2003; Nemitz et al., 2002; Grimmond et al., 2002), as well as
aerosols (Dorsey et al., 2002). Measurements of momentum and heat fluxes provide basic data
to run air quality models and valuable information for understanding the dispersion of pollutants
in the urban atmosphere. In particular, measurement of heat flux is important for determining
the intensity of an urban heat island. An urban heat island influences urban ecology in a variety
of ways by altering such things as the physiological comfort of humans, cooling requirements,
duration of snow cover, cloud cover, precipitation rates, etc. CO2 flux measurements have
12
validated urban emissions inventories and have shown that emission patterns depend on the
diurnal cycle and climatic season. In general, urban landscapes are net sources of CO2, and their
emission magnitude depends on climatic conditions (e.g. temperature and humidity), fraction of
land covered by vegetation, vehicular traffic, population density, etc. The only study reporting
aerosol fluxes is that by Dorsey et al. (2002) who related aerosol flux with vehicular traffic and
boundary layer stability in the City of Edinburgh.
1.5. VOCs: photochemical precursors and toxic pollutants
By definition the term VOC is used to denote all organic compounds that have a vapor
pressure higher than 13.3 Pa at 25°C, according to ASTN test method D3960–90. However, in
terms of air quality, VOCs encompass all of the individual reactive organics in the troposphere
that contribute to the creation of photochemical smog and tropospheric ozone.
VOCs are a key piece in understanding photochemical air quality in urban atmospheres. In
the presence of sunlight and NOx, VOCs are oxidized to carbon dioxide and water via various
intermediates: radicals (e.g., hydroxyl, hydroperoxy, organic peroxy), organics (e.g., aldehydes,
acids, alcohols, nitrates, peroxides), and inorganics (e.g., carbon monoxide, ozone, hydrogen
peroxide, nitric acid) (Finlayson-Pitts and Pitts, 1997). It has been demonstrated that many of
these secondary compounds may have direct health impacts (Evans et al., 2002), and that some
individual VOC are extremely toxic pollutants (e.g., the carcinogens benzene and 1-3-butadiene).
Therefore, accurate and precise measurements of VOCs are mandatory for improving the air
quality in urban areas, and especially in megacities, where the highest concentrations of ozone
and VOCs have been reported (Molina and Molina, 2004; Barletta et al., 2002).
The almost universal approach for atmospheric hydrocarbon measurements is based upon gas
chromatographic (GC) separation of the individual hydrocarbons and the detection of each
compound using either a flame ionization detector (FID) or a mass spectrometer (MS). GC-MS
is used to establish the identity of a particular compound through the combination of retention
times and mass spectra and, of course, can also be used for quantification. However, GC-FID is
13
commonly used for more extensive quantitative measurements after the individual peaks have
been identified (Parrish and Fehsenfeld, 2000). Collection of air samples in stainless steel
canisters is the most common collection technique for VOCs (Westberg and Zimmerman, 1993).
New techniques are emerging that offer relatively fast response and continuous
measurements of certain hydrocarbons. One method, designated here as the fast olefin sensor, is
based on the chemiluminescent reaction between olefins and ozone (Guenther and Hills, 1998).
Originally it was designed to measure isoprene for EC studies in natural landscapes. Recently, it
has been applied in urban areas to measure the olefinic mix, since it has a significant response to
alkenes including propylene, ethylene, and 1,3-butadiene, among others. Another novel
technique is the Proton Transfer Reaction Mass Spectrometry, which is capable of accurate,
selective and fast-response measurements of a number of oxygenated and aromatic
hydrocarbons. This technique is based upon the property that proton transfer from H3O+ ions can
ionize many hydrocarbons, usually without fragmenting the parent ion (Lindinger et al., 1998).
Optical techniques are a third method to measure ambient VOC. The three most common
optical methods are: DOAS, FTIR, Tunable Diode Laser Spectroscopy (TDLAS). The DOAS
technique is based on the UV-visible molecular absorption of atmospheric gases. It offers the
possibility of measuring several VOCs simultaneously. It provides the average concentration
over a long optical path (from 100 m to 10 km), thus avoiding local perturbations that can
compromise point measurements. DOAS has been developed to the point that BTEX (benzene,
toluene, ethylbenzene, xylenes) and other monocyclic aromatic hydrocarbons can be monitored
with sensitivities of ppbv on minute time scales.
Air samples can contain hundreds of different hydrocarbons of natural and anthropogenic
origin; for example, over 850 different hydrocarbons have been detected in the vapor over
gasoline, and over 300 different hydrocarbons from a vehicle exhaust-polluted air sample have
been identified (Parrish and Fehsenfeld, 2000). Not all VOCs have the same reactivity, or rather
the same impact on the formation of ozone and other secondary pollutants. The relative
reactivity of individual VOCs can differ by more than one order of magnitude (Rusell et al.,
14
1995). Ignoring the reactivity of emissions when regulations are developed may lead to
ineffective, inefficient control strategies and possibly even lead to measures that worsen air
quality.
1.6. CO2, a greenhouse gas
Carbon dioxide is a greenhouse gas that absorbs infrared photons and reemits radiation in all
directions, including back towards the Earth’s surface. This additional back-radiation, from all
greenhouse gases combined, makes the surface-air temperature about 33ºC warmer than it would
be in the absence of such radiation. Water vapor is the most dominant greenhouse gas; CO2 is
second (Finlayson-Pitts and Pitts, 2000).
Fossil-fuel combustion adds CO2 to the atmosphere, and it has been recognized that this
additional source is contributing CO2 faster than the oceans and terrestrial biosphere can remove
it. Therefore, CO2 is accumulating in the atmosphere and enhancing the greenhouse effect.
Since the industrial revolution, the CO2 concentration in the atmosphere has increased from
approximately 280 ppmv to greater than 360 ppmv today. It is expected that atmospheric CO2
levels will continue to rise, and during the century could well exceed 500 ppmv (IPCC, 2001).
This rise would lead to a number of severe environmental and economic consequences: changes
in frequency, intensity, and duration of extreme events, such as more hot days, heat waves, heavy
precipitation events, and fewer cold days; an increase in threats to human health, predominantly
within tropical and subtropical countries; crop yields could be adversely affected; and water
shortages in many water-scarce areas of the world will be exacerbated (IPCC, 2001).
All urban areas, large and small, in developing and developed areas of the world have been
identified as major sources of anthropogenic CO2 (Decker et al., 2000). Many methodologies
have been developed to estimate urban CO2 emissions. Usually they are based on fuel
inventories or by other indirect means that involve a number of uncertainties or require
information that may not be available or reliable for cities in less developed countries. Direct
measurements using micrometeorological techniques have only recently been applied in a few
15
cities (Soegaard and Møller-Jensen, 2003; Moriwaki and Kanda, 2004; Grimmond et al., 2004).
Despite the fact that three-quarters of the world’s 3 billion urban residents live in developing
countries, few, if any, flux measurements have been made in these urban areas. For these cities,
emission inventories will have to be built by combining micrometeorological techniques, remote
sensing, and conventional methodologies.
1.7. References
Arriaga-Colina, J.L., West, J.J., Sosa, G., Escalona, S.S., Orduñez, R.M. & Cervantes, A.D.M. Measurements of VOCs in Mexico City (1992-2001) and evaluation of VOCs and CO in the emissions inventory. Atmospheric Environment 38, 2523-2533 (2004).
Barletta, B., et al. Mixing ratios of volatile organic compounds (VOCs) in the atmosphere of Karachi, Pakistan. Atmospheric Environment 36, 3429-3442 (2002).
Bossert, J.E. An investigation of flow regimes affecting the Mexico City region. J. Applied Meteorology 36, 119-140 (1997).
Chameides, W.L., Kasibhatla, P.S., Yienger, J. & Levy II, H. Growth of continental-scale metro agro-plexes, regional ozone pollution, and world food production. Science 264, 74-77 (1994).
Decker, E.H., Elliot, S., Smith, F.A., Blake, D.R. & Rowland, F.S. Energy and material flow through the urban ecosystem. Annu. Energy Environ. 25, 685-740 (2000).
Dockery, D.W., et al. An association between air pollution and mortality in six US cities. New England Journal of Medicine 329(24), 1753-1759 (1993).
Dorsey, J.R. et al. Direct measurements and parameterization of aerosol flux, concentration and emission velocity above a city. Atmospheric Environment 36, 791-800 (2002).
Evans, J., et al. in Quality in the Mexico Megacity: an integrated assessment (ed. Molina, L.T. & Molina M.J.) 105-136 (Kluwer Academic Publishers, Netherlands, 2002).
Finlayson-Pitts, B.J. & Pitts, J.N. Chemistry of the upper and lower atmosphere. (Academic Press, USA 2000).
Fujita, E.M., et al. Comparison of emission and ambient concentration ratios of CO, NOx, and NMOG in California’s south coast air basin. Journal of Air and Waste Management Association 42, 264-276 (1992).
Gauss, M., et al. Radiative forcing in the 21st century due to ozone changes in the troposphere and the lower stratosphere. Journal of Geophysical Research 108, 4292 (2003).
16
Gobierno del Distrito Federal (GDF). Informe del Estado de la Calidad del Aire y Tendencias 2003 para la Zona Metropolitana del Valle de México. (Dirección General de Gestión Ambiental del Aire. México, D.F., 2004).
Goldan, P.D. et al. Measurements of hydrocarbons, oxygenated hydrocarbons, carbon monoxide, and nitrogen oxides in an urban basin in Colorado: implications for emission inventories. Journal of Geophysical Research 100, 22771-22783 (1995).
Grimmond, C.S.B., King, T.S., Cropley, F.D., Nowak, D.J. & Souch, C. Local-scale fluxes of carbon dioxide in urban environments: methodological challenges and results from Chicago. Environmental Pollution 116, 243–254 (2002).
Grimmond, C.S.B., Salmond J.A., Oke, T.R., Offerle, B. & Lemonsu, A. Flux and turbulence measurements at a densely built-up site in Marseille: heat, mass (water and carbon dioxide), and momentum. Journal of Geophysical Research 109, 24101-24120 (2004).
Guenther, A. & Hills, A. Eddy covariance measurement of isoprene fluxes. Journal of Geophysical Research 103, 13145-13152 (1998).
Guenther, A. et al. Isoprene fluxes measured by enclosure, relaxed eddy accumulation, surface layer gradient, mixed layer gradient, and mixed layer mass balance techniques. Journal of Geophysical Research 101, 18555-18567 (1996).
Gurjar, B.R. & Lelieveld, J. New directions: megacities and global change. Atmospheric Environment 39, 391-393 (2005).
Intergovernmental Panel on Climate Change (IPCC). Climate Change 2001: The scientific basis. Contribution of working group I to the third assessment report of the intergovernmental panel on climate change. (Cambridge University Press, Cambridege, UK, 2001).
Jacob, D.J., Logan, J.A. & Murti, P.P. Effect of rising Asian emissions on surface ozone in the United States. Geophysical Research Letters 26, 2175-2178 (1999).
Karl, T.G. et al. Virtual disjunct eddy covariance measurements of organic compound fluxes from a subalpine forest using proton transfer reaction mass spectroscopy. Atmospheric Chemistry and Physics 2, 279-291 (2002).
Kiehl, J.T., Schneider, T.L., Portmann, R.W. & Solomon, S. Climate forcing due to tropospheric and stratospheric ozone. Journal of Geophysical Research 104, 31239-31254 (1999).
Lindinger, W., Hansel, A. & Jordan, A. On-line monitoring of volatile organic compounds at pptv levels by means of proton-transfer-reaction mass spectroscopy (PTR-MS): Medical applications, food control and environmental research. International Journal of Mass Spectrometry Ion Processes 173, 191-241 (1998).
McGaughey, G.R., Desai, N.R., Allen, D.T., Seila, R.L., Lonneman, W.A., Fraser, M.P., Harley, R.A, Pollack, A.K., Ivy, J.M. & Price, J.H. Analysis of motor vehicle emissions in a Houston
17
tunnel during the Texas Air Quality Study 2000. Atmospheric Environment 38, 3363-3372 (2004).
Molina, L.T., Molina, M., Favela, R., Fernandez, A., Slott, R. & Zavala, M. in Quality in the Mexico Megacity: an integrated assessment (ed. Molina, L.T. & Molina M.J.) 21-55(Kluwer Academic Publishers, Netherlands, 2002).
Molina, M.J. & Molina L.T. Megacities and atmospheric pollution. Air and Waste Management Association 54, 644-680 (2004).
Molina, M.J., Molina, L.T., West J., Sosa G., Sheibaum C. in Quality in the Mexico Megacity: an integrated assessment (ed. Molina, L.T. & Molina M.J.) 137-202 (Kluwer Academic Publishers, Netherlands, 2002).
Moriwaki, R. & Kanda, M. Seasonal and diurnal fluxes of radiation, heat, water vapor, and carbon dioxide over a suburban area. Journal of Applied Meteorology 43, 1700-1710 (2004).
Nemitz, E., Hargreaves, K.J., Mcdonald, A.G., Dorsey, J.R. & Fowler, D. Micrometeorological measurements of the urban heat budget and CO2 emissions on a city scale. Environmental Science and Technology 36(14), 3139-3146 (2002).
Parrish, D.D. & Fehsenfeld, F.C. Methods for gas-phase measurements of ozone, ozone precursors and aerosol precursors. Atmospheric Environment 34, 1921-1957 (2000).
Pressley, S., Lamb, B., Westberg, H., Flaherty, J. & Chen, J. Long-term isoprene mesurements above a northern hardwood forest. Journal of Geophysical Research 110, doi:10.1029/2004JD005523 (2005).
Prinn, R.G., et al. Evidence for substantial variations of atmospheric hydroxyl radicals in the past two decades. Science 282, 1882-1998 (2001).
Rinne, H.J.I, Guenther, A., Warneke, C., de Gouw, J.A. & Luxembourg, S.L. Geophysical Research Letters 28, 3139-3142 (2001).
Rusell, A., et al. Urbane ozone control and atmospheric reactivity of organic gases. Science 269, 491-495 (1995).
Sawyer, R.F. et al. Mobile sources critical review : 1998 NARSTO assessment. Atmospheric Environment 34, 2161-2181 (2000).
Secretaria del Medio Ambiente del Gobierno del Distrito Federal (SMA-GDF). Inventario de emisiones de la Zona Metropolitana del Valle de México (Draft). Available at: http://www.sma.df.gob.mx (2005).
Shaw, W.J., Spicer, C.W. & Kenny, D.V. Eddy correlation fluxes of trace gases using a tandem mass spectrometer. Atmospheric Environment 32, 2887-2898 (1998).
18
Sillman, S. The relation between ozone, NOx, and hydrocarbons in urban and polluted rural environments. Atmospheric Environment 33, 1821-1845 (1999).
Singer, B.C. & Harley, R.A. A fuel-based motor vehicle emission inventory. Journal of the Air and Waste Management Association 46, 581–593 (1996).
Soegaard, H. & Møller-Jensen, L. Towards a spatial CO2 budget of a metropolitan region based on textural image classification and flux measurements. Remote Sensing of Environment 87, 283-294 (2003).
Spirig, C., Neftel, A., Ammann, C., Dommen, J., Grabmer, W., Thielmann, A., Schaub, A., Beauchamp, J., Wisthaler, A. & Hansel, A. Eddy covariance flux measurements of biogenic VOCs during ECHO 2003 using proton transfer reaction mass spectrometry. Atmospheric Chemistry and Physics 5, 465–481 (2005).
Trainer, M., Parrish, D.D., Goldan, P.D., Roberts, J. & Fehsenfeld, F.C. Review of observation-based analysis of the regional factors influencing ozone concentrations. Atmospheric Environment 34, 2045–2061 (2000).
United Nations (UN). World Urbanization Prospects. The 2003 Revision. Avialable at: http://www.un.org/esa/population/publications/wup2003/2003WUPHighlights.pdf (2004).
Velasco, E., Lamb, B., Pressley, S.., Allwine, E., Westberg, H., Jobson, B.T., Alexander, M., Prazeller, P., Molina, L. & Molina, M. Flux measurements of volatile organic compounds from an urban landscape. Geophysical Research Letters 32(20), L20802, doi:10.1029/2005GL023356 (2005a).
Velasco, E., Pressley, S., Allwine, E., Westberg, H., Lamb, B. Measurements of CO2 fluxes from the Mexico City urban landscape. Atmospheric Environment 39(38), 7433-7446 (2005b).
Watson, J.G., Chow, J.C. & Fujita, E. Review of volatile organic compound source apportionment by chemical mass balance. Atmospheric Environment 35, 1567-1584 (2001).
West, J.J. et al. Modeling ozone photochemistry and evaluation of hydrocarbon emissions in the Mexico City Metropolitan Area. Journal of Geophysical Research 109, 19312 (2004).
Westberg, H. & Zimmerman, P. Analytical methods used to identify nonmethane organic compounds in ambient atmospheres. Advances in Chemistry Series 232, 275-290 (1993).
Westberg, H. et al. Measurement of isoprene fluxes at the PROPHET site. Journal of Geophysical Research 106, 24347-24358 (2001).
19
20
CHAPTER 2
DISTRIBUTION, MAGNITUDES, REACTIVITIES, RATIOS AND DIURNAL
PATTERNS OF VOLATILE ORGANIC COMPOUNDS IN THE VALLEY OF MEXICO
DURING THE MCMA 2002 & 2003 FIELD CAMPAIGNS
2.1. Abstract
A wide array of volatile organic compound (VOC) measurements was conducted in the
Valley of Mexico during the MCMA-2002 and 2003 field campaigns. Study sites included
locations in the urban core in, a heavily industrial area and at boundary sites in rural landscapes.
Four distinct analytical techniques were used: whole air canister samples with gas
chromatography / flame ionization detection (GC-FID), on-line chemical ionization using a
Proton Transfer Reaction Mass Spectrometer (PTR-MS), continuous real-time detection of
olefins using a Fast Olefin Sensor (FOS), and long path measurements using UV differential
optical absorption spectrometers (DOAS). The simultaneous use of these techniques provided a
wide range of individual VOC measurements with different spatial and temporal scales. The
VOC data were analyzed to understand concentration and spatial distributions, diurnal patterns,
origin and reactivity in the atmosphere of Mexico City. On the basis of concentrations given in
ppbC, the VOC burden was dominated by alkanes (60%), followed by aromatics (15%) and
olefins (5%). The remaining 20% was a mix of alkynes, halogenated hydrocarbons, oxygenated
species (esters, ethers, etc.) and other unidentified VOCs. However, in terms of ozone
production, olefins were the most relevant hydrocarbons. Elevated levels of toxic hydrocarbons,
such as 1,3-butadiene, benzene, toluene and xylenes were also observed. Various analytical
techniques have demonstrated that vehicle exhaust is the main source of VOCs in urban areas
and that diurnal patterns depend on vehicular traffic. Finally, examination of the VOC data in
terms of lumped modeling VOC classes and comparison to the VOC emissions reported in the
21
available Mexican inventory derived for purposes of photochemical air quality suggests that
some, but not all, classes are underestimated in the inventory by factors of 2 or 3.
2.2. Introduction
Volatile Organic Compounds (VOCs) play a key role in photochemical air quality in urban
atmospheres. In the presence of sunlight and nitrogen oxides (NOx), VOC oxidation produces
secondary products including radicals (e.g., hydroxyl, hydroperoxy, organoperoxy), oxygenated
organics (e.g., aldehydes, ketones, acids, nitrates, peroxides), and inorganics (e.g., carbon
monoxide, ozone, hydrogen peroxide, nitric acid) (Finlayson-Pitts and Pitts, 1997). It has been
demonstrated that many of these secondary compounds may have direct health impacts (Evans et
al., 2002), and that some individual VOCs are extremely toxic pollutants (e.g., the carcinogens
benzene and 1-3-butadiene). In a short-term study such as reported in this manuscript, it is not
possible to quantify the health impacts from exposure to air toxic species because health
benchmarks are measured in terms of annual averages. Long term measurements of VOCs are
needed for understanding air toxic issues in urban areas, especially in megacities such as Mexico
City, where millions of people are exposed to high concentrations of VOCs and other pollutants
(Molina and Molina, 2004).
Mexico City with a population of 19 million, is a good example of a megacity with severe
air pollution problems. It is situated in the Valley of Mexico and is the capital city of a
developing country. Mexico City is in the subtropical zone as are most of the world’s fastest
growing cities. Many of the most populated Latin-American urban areas, like Mexico City are at
high altitude which makes combustion processes less efficient leading to enhanced VOC
emissions. Mexico City’s high altitude (~2240 m) and low latitude (19o 25’ N), subtropical
weather and intense solar radiation accentuate the VOC evaporative emissions from a variety of
sources such as storage and distribution of gasoline, cleaning, painting, and industrial processes.
However the main contributors to high VOC concentrations in Mexico City are the extensive
presence of aged industrial operations (more than 53,000 industries) and a relatively old vehicle
22
fleet (more than 3.5 million vehicles with an average age of ~9 years). The elevated
anthropogenic emissions, the intense solar radiation and area’s topography with mountains to the
west, east and south of the valley (see Figure 1) produce elevated levels of photochemical
pollutants that are trapped in the basin on a daily basis.
Many researchers have addressed the VOC pollution problem in Mexico City. Ambient
VOC concentrations have been evaluated since the early 90’s (see papers cited by Raga et al.,
2001). All studies have consistently reported high concentrations of propane, butane and other
low molecular weight alkanes, which have been attributed to liquefied petroleum gas (LPG)
leakage during handling, distribution and storage. LPG is the main fuel for cooking and water
heating in Mexican households. High ambient concentrations of photochemical reactive VOCs,
such as olefins and aromatics have been reported, as well (Mugica et al., 2002; Arriaga et al.
1997). These two VOC groups have been identified to be the main species responsible for the
ozone and secondary organic aerosol formation in Mexico City (Gasca et al., 2004; Mugica et
al., 2002). High exposure levels to aromatics hydrocarbons, particularly to toluene, benzene and
xylenes have been detected in different microenvironments and public transport of Mexico City
(Shiohara et al., 2005, Gomez-Perales et al., 2004; Cruz-Nuñez et al., 2003; Bravo et al., 2002;
Ortiz et al., 2002; Meneses et al., 1999).
From source apportionment studies, it has been determined that motor-vehicles, especially
gasoline-vehicles are the main source of those aromatic hydrocarbons (Mugica et al., 2003).
Emissions profiles for motor-vehicles in on-road conditions have been obtained from tunnel and
remote sensing studies (Schifter et al., 2003; Mugica et al., 1998). However, many uncertainties
exist about vehicular emissions in Mexico City as consequence of the lack of reliable emission
factors and daily activity levels of the fleet (Gakenheimer et al., 2002). VOC emission profiles
are also available for food cooking, asphalt application and painting operations, among other
types of sources in Mexico City (Vega et al., 2000; Mugica et al. 2001). The biogenic
component in the VOC emission burden has been estimated to contribute no more than 7% in the
Valley of Mexico (Velasco, 2003).
23
During the last decade, a number of policies and actions have been enacted to decrease VOC
emissions in Mexico City, among them, the enforcement of vapor recovery systems in fuel
service stations, the banning of leaded gasoline and heavy oil fuels, the availability of diesel with
reduced sulfur content and the substantial reduction of olefins, benzene, and other aromatic
hydrocarbons content in gasoline. Arriaga-Colina et al. (2004) reported that, as result of these
emission control measures, ambient VOC concentrations have stabilized and possibly decreased,
despite the growth in the vehicular fleet and other activities. However, ambient VOC levels still
remain high and the national air quality standard for ozone (110 ppb/1 hour average) is exceeded
70% of the days (GDF, 2004).
Although some advances have been achieved, it is clear that a better understanding of the
photochemical pollution in the complex urban ecosystem of Mexico City is needed to support
effective control strategies. In this context, a number of US and Mexican institutions and
agencies participated in the Mexico City Metropolitan Area 2002 and 2003 (MCMA-2002 &
MCMA-2003) field campaigns. The first was an exploratory campaign performed in February
2002, and the second an intensive five week field study during April and May, 2003. The goals
of both field studies were to update and improve the emissions inventory of Mexico City, and to
gain a better understanding of the chemistry and transport processes driving atmospheric
pollution in the valley. As part of the investigation to meet these goals, a wide array of VOC
measurements was conducted at airshed boundary sites, central urban core sites, and downwind
urban receptor sites. Four different analytical methods were used: VOC speciation by gas
chromatographic analysis using flame ionization detection (GC-FID), olefin detection with a Fast
Olefin Sensor (FOS), determination of a number of oxygenated and aromatic VOCs by Proton
Reaction Transfer Mass Spectroscopy (PTR-MS), and measurements of selected VOCs using
Differential Optical Absorption Spectroscopy (DOAS). Key aspects of these VOC
measurements were the number of individual species measured and the variety of spatial and
temporal scales employed in the measurements.
24
This manuscript presents a summary of results from the different VOC measurements in
terms of the distribution, magnitudes, and diurnal patterns of selected VOCs. The use of ratios
of individual VOC species is used to characterize different sites, to investigate the relative
reactivity of different species and to determine their origins through comparisons with source
signatures, in particular with vehicle exhaust profiles. In addition, the ambient VOC data has
been compared to the lumped VOC emissions classes in the Mexican emissions inventory used
for photochemical air quality modeling. It has been suggested that VOC emissions are
underestimated in the official emissions inventory by a factor of 3 from analysis of the
VOC/NOx ratio (Arriaga-Colina et al., 2004) and from ozone modeling exercises (West et al.,
2004). However, eddy covariance flux measurements of selected VOCs during the MCMA-2003
campaign suggest that for the VOC classes measured, the VOC emissions reported in the
emissions inventory are generally correct (Velasco et al., 2005).
2.3. Monitoring sites
During the MCMA-2002 field campaign only instantaneous canister samples were collected
from five different sites. For the MCMA-2003 study, new sites were included and diverse
measurement techniques were implemented. Figure 1 shows a map of the Valley of Mexico
indicating the VOC sites in both campaigns. Overall, eight sites were employed, four
corresponded to urban sites with different characteristics (La Merced, Constituyentes, Pedregal
and CENICA), one to an industrial section of the city (Xalostoc), and three to boundary sites in
rural areas (La Reforma, Teotihuacan and Santa Ana Tlacotenco). Table 1 lists a summary of
these sites and VOC measurement methods used during the MCMA-2002 and MCMA-2003
field campaigns.
25
2.4. Instrumentation
2.4.1. Gas chromatography separation and flame ionization detection (GC-FID)
Ambient VOC samples were collected from all monitoring sites in Summa® electro-polished
stainless-steel canisters. During the MCMA-2002 study, 46 samples were filled instantaneously,
while for the MCMA-2003 study, 184 samples were collected with (usually) a three hour
averaging interval using automated samplers. From both studies, 64% of the samples were
collected during the morning rush traffic period (6:00 to 9:00 h) when VOC concentrations are
strongly related to anthropogenic emissions and before the photochemical processing has begun.
The remaining samples were collected during the late morning and early afternoon hours.
Samples were collected and analyzed by two different research groups: Washington State
University (WSU) and the Mexican Petroleum Institute (IMP). WSU collected and analyzed
78% of the samples. During the MCMA-2003 study, WSU collected all its samples with a
XonTech, Inc. Air Sampler model 910PC. Half of the samples were analyzed on-site within 24
hours of collection to minimize storage losses before analysis. IMP collected samples only in
2003 using an AVOCS Anderson Automated Sampler with a Viton diaphragm pump. Both
groups analyzed their samples using methodology similar to the US EPA compendium methods
TO-14/15 (US-EPA, 1999). The GC analysis utilized cryogenic pre-concentration by drawing
air from the canisters through a stainless-steel loop containing glass beads (60/80 mesh) and
cooled to liquid oxygen temperature. WSU employed a Hewlett-Packard 6890 Series
chromatograph. The GC was equipped with a 30-m fused silica DB-1 column (0.32 mm i.d. and
1 µm film thickness) with a 2 ml min-1 carrier flow. Prior to sample injection, the oven was
cooled to -50ºC. During analysis, the oven temperature was raised at 4ºC min-1 to a final
temperature of 150ºC. The total analysis time was approximately one hour. Detector response
was calibrated with NIST traceable 2,2-dimethylbutane standard. A minimum detection limit of
20 pptC was determined. Individual species were identified by retention times.
IMP conducted their analysis using a Hewlett-Packard 5890 Series II chromatograph
containing a 60-m Quadrex fused silica glass capillary column (0.32 mm i.d. and coated with a 1
26
µm film thickness) at a flow of 2 ml min-1. The oven temperature started at -50ºC and was
heated up to 200ºC at a rate of 8ºC min-1. The FID response was calibrated with a certified high-
purity propane standard from Praxair. The detection limit was determined to be 1 ppbC.
Individual species were identified by retention times using a mixture of 55 hydrocarbons (Scott
Specialty Gases NIST Traceable), and a certified mixture of 33 halogen-containing compounds
(Spectra Gases, with 10% analytical accuracy).
2.4.2. The Proton Transfer Reaction Mass Spectrometry (PTR-MS)
The Proton Transfer Reaction Mass Spectrometry identifies VOCs in ambient air as their
molecular mass plus one. This technique creates ions by transferring a H+ from H3O+ to the
VOCs followed by mass spectroscopy detection of the product ions (Lindinger et al., 1998). The
PTR-MS does not employ a column, so response times are short (seconds) and automated,
continuous measurements can be made over extended periods of time. Specificity in the PTR-
MS is achieved by the soft ionization (minimal fragmentation) and response is limited to species
with a greater proton affinity than water. In cases where several VOCs produce the same M+1
ion, individual species quantitation is not possible. For example, the signal at mass 121 (C3-
benzenes) is comprised of i and n-propylbenzene, three ethyltoluene and three trimethylbenzenes
isomers. Validation of PTR-MS measurements have been performed by de Gouw et al. (2003)
and Warnake et al. (2003) to determine the set of VOCs that are suitable for measurements with
this technique.
During the MCMA-2003 field campaign, two PTR-MS instruments were used. One was
operated at the CENICA site by the Pacific Northwest National Laboratory (PNNL), and the
second PTR-MS instrument, belonging to Montana State University (MSU), was housed in a
mobile laboratory used primarily for vehicle chase experiments. During selected periods the
mobile laboratory was employed as a fixed monitoring station at La Merced, Santa Ana
Tlacotenco and Pedregal sites. This manuscript presents only PTR-MS fixed-site measurements.
The species monitored by PTR-MS included benzene, toluene, styrene, C2-benzenes, C3-
27
benzenes, naphthalene, phenol, cresols, methanol, acetaldehyde, acetone and acetonitrile. The
C2-benzenes are represented by mass 107, which has contributions from ethylbenzene, o,m & p-
xylene, and benzaldehyde (de Gouw et al., 2003).
Both PTR-MS instruments were calibrated in the field using a multi-component gas standard
that contained the species reported here. The calibration standard was diluted with humidified
zero air in order to generate a multipoint calibration curve over the 1 to 50 ppbv range.
Calibrations were performed every 2–3 days. The instrument background was automatically
recorded every 3 hours by switching the sample flow to a humid zero air stream. Zero air was
continuously generated by passing ambient air through a Pt-catalyst trap heated to 300 ˚C.
Background count rates were subtracted from the ambient data.
2.4.3. Fast Olefin Sensor (FOS)
Continuous real-time measurements of olefin concentrations were made at the CENICA site
during the MCMA-2003 study by WSU using a FOS. The FOS is a fast isoprene sensor
(Guenther and Hills, 1998) based on chemiluminescent that occurs when an olefinic bond reacts
with ozone. The chemiluminescent response varies considerably for individual olefins. In an
urban atmosphere where numerous olefins are present, it is necessary to evaluate the FOS
response to as many olefins as possible. For Mexico City, we determined response
characteristics for five olefins (propylene, ethylene, isoprene, 1-butene and 1-3 butadiene). Since
NO levels are known to be high in the Mexico City atmosphere, we investigated potential
interferences due to its reaction with ozone. Nitric oxide gave no response in this test. Response
results for the olefins are shown in Table 2 along with their corresponding relative sensitivities to
propylene. It was found that the FOS is more sensitive to 1-3-butadiene and isoprene than to
propylene; however, their ambient concentrations were much lower than the ambient levels of
propylene. In contrast, species with a lower sensitivity than propylene, but with high
concentrations in urban atmospheres (e.g. ethylene) can produce large signals.
28
During the campaign, the FOS was calibrated 3 times per day using dilutions from a
propylene standard (Scott Specialty Gases, 10.2 ppm, ±5% certified accuracy). The linear slope
of instrument response versus propylene concentration and the zero level exhibited relatively
little drift during the study period, 14% and 9%, respectively.
2.4.4. Differential Optical Absorption Spectrometry (DOAS)
During the MCMA-2003 field campaign, three DOAS instruments were employed to
measure ambient VOC concentration. DOAS is based on the UV-molecular absorption of
atmospheric gases and measures continuously concentrations of a number of trace gases
averaged over a long optical path. The Massachusetts Institute of Technology (MIT) deployed a
DOAS system 16 m above the surface with a path length of 960 m to measure benzene, styrene,
m-xylene, p-xylene, ethylbenzene, benzaldehyde, phenol, cresols, naphthalene and formaldehyde
at the CENICA site. A second DOAS (L = 860 m, H = 70 m) was employed by the MIT group
to measure special trace gases such as glyoxal and HONO. A third DOAS system (L = 426 m, H
= 20 m) was deployed at La Merced by the National Autonomous University of Mexico
(UNAM) to measure benzene and toluene.
2.5. Ambient air inter-comparison of GC, PTR-MS and DOAS measurements for selected
VOCs
Table 1 shows that at 50% of the monitored sites, two or more different techniques were
used to measure VOCs. This provides the opportunity to inter-compare data to verify that they
yield consistent results. Since no single technique can provide speciated measurements of all the
VOCs needed for photochemical modeling, different measurement techniques are often used in
combination to provide information for as many species as possible.
Figures 2a and 2b show time series of benzene and toluene mixing ratios measured by GC-
FID (WSU), PTR-MS (MSU) and DOAS (UNAM) at the La Merced site. Both figures show
generally good temporal agreement for benzene and toluene. Some differences between
29
concentration levels measured by DOAS and the other two methods are to be expected. The
DOAS signal represents average concentrations over an open path distance of about 426 m while
the PTR-MS and GC data are from measurements at a specific location. The PTR-MS technique
is the most sensitive to transient plumes from close-by sources because it measures in real-time
with no averaging. This feature can be seen in Figure 2a where benzene concentrations
determined by the PTR-MS are much more variable than those from DOAS or GC, for example
the hourly standard deviations of concentrations measured by the PTR-MS are twice those
measured by DOAS. Furthermore, it is evident in this figure that the GC results agree better
with the PTR-MS than with DOAS. Benzene levels measured by DOAS appear to exceed the
concentrations determined by the other two techniques, especially during the afternoon hours.
This may indicate that the point measurements (GC and PTR-MS) underestimate the regional
benzene abundance and/or the DOAS measurement of benzene is subject to an interference
during the midday period. It’s unusual to see a benzene peak in the 10-16:00 time period.
The toluene time series shows better agreement between techniques and a more normal
urban pattern of high nighttime and early morning concentrations and much lower levels during
the midday period. Jobson et al. (2005) present a more detailed inter-comparison with the same
three techniques at the CENICA site. They found that the level of agreement between point and
long path techniques is influenced by wind direction, but in general terms they also found
relatively good agreement between GC, PTR-MS and DOAS techniques.
Figure 3 shows time series plots of C2-benzenes and C3-benzenes measured by the MSU
PTR-MS together with GC-FID samples collected during the same time periods at the Pedregal,
Santa Ana Tlacotenco and La Merced sites. At La Merced and Pedregal sites, good agreement
between methods for the two aromatic groups was observed. Temporal variability correlate well
and GC-FID mixing ratios were always within the one standard deviation range of the PTR-MS
measurements. Agreement was not as good at the Santa Ana Tlacotenco site. GC-FID C3-
benzenes concentrations were quite often above the one standard deviation range of the PTR-MS
concentrations. Note that concentrations at this site were one order of magnitude smaller than
30
concentrations at the other sites. Benzene agreement between GC and PTR-MS was not as good
at the rural Santa Ana Tlacotenco site as at the urban sites. GC-FID measurements of benzene
were always near the lower limit of the one standard deviation range of the PTR-MS
concentrations. However, toluene concentrations agreed closely at all three sites.
Analytical consistency was examined for the two GC techniques by collecting parallel
samples using WSU and IMP samplers. WSU reported 57 hydrocarbons up to 10 carbons, while
IMP quantified 104 VOCs up to 12 carbons, including oxygenated and halogenated species. The
extra 47 species determined by IMP represented less than 10% of the total VOC burden.
Halogenated species comprised about 2.5% of the total VOC total. Analytical results for the
each of the 57 compounds directly compared during four different sampling periods showed a
VOCWSU/VOCIMP ratio between 0.9 and 1.10.
A comparison of the FOS response to the sum of olefins as measured simultaneously with
the canister sampling system suggests that the total olefin level detected by the FOS is larger
than the sum of identified olefins from canister samples. The ratio between the sum of olefins
measured by canisters and the FOS signal showed a median of 48%. This suggests that 52% of
olefins detected by the FOS remain unknown or that the use of propylene as the calibration
standard does not adequately represent the urban olefin mix. Additional laboratory studies are
needed resolve this uncertainty. However, it can be affirmed that the FOS measures a mix of
VOCs responding as propylene and can be used to provide a continuous and fast response
measurement of the olefinic VOC mix in an urban atmosphere.
2.6. Results and discussion
2.6.1. Diurnal patterns of various VOCs at selected sites
Although the VOC time series shown in Figures 2 and 3 only cover a few days, they provide
insight concerning the diurnal pattern of VOCs at sites with different characteristics in the Valley
of Mexico. For example, at the La Merced site, concentrations of benzene, toluene, C2-benzenes
and C3-benzenes reach their highest level during the morning rush hour at 7:00 h. The
31
accumulation of these species in the surface-based layer starts in the evening of the previous day.
During the night the stability of the surface-based layer increases, and trapps pollutants emitted
near the surface. Near sunrise, the VOC concentrations increase due to the beginning of the
daily vehicular traffic. Between 10 and 11:00 h, concentrations decrease significantly because
the nighttime surface inversion erodes with consequent growth of the boundary layer. The
typical mixing height changes from less than 500 m in the early morning to heights above 2000
m at noon (Whiteman et al., 2000).
It is interesting that no late afternoon peak in VOC concentration was observed at the La
Merced site, in contrast to the Pedregal site, where morning and late afternoon VOC peaks were
observed. The late peak at Pedregal was observed to occur during the evening between 18 and
22:00 h. This difference between the La Merced and Pedregal sites appears to be due to different
diurnal traffic patterns. Traffic counts performed during the MCMA-2003 campaign on avenues
close to the two sites indicated that at La Merced had a single traffic peak that begins at 6:00 h
and extends until late evening. On avenues near the Pedregal site, the typical morning and
afternoon traffic peaks could be identified. Note that VOC concentrations were higher at La
Merced than at Pedregal by a factor of ~3. The La Merced site was located immediately next to
a number of avenues, while the Pedregal site was more distant from main roadways.
Santa Ana Tlacotenco is a rural site located in a mountain gap in the southeast corner of the
Valley of Mexico through which pollutants are advected to the contiguous Valley of Cuautla.
This drainage usually occurs during nighttime and the morning. During afternoons and early
evenings air flow through this gap is in the opposite direction. The air coming back into the
Valley of Mexico may contain diluted concentrations of pollutants that originated there at an
earlier time. This drainage effect has been described in detail by Doran and Zhong (2000) and
Jazcilevich et al. (2003). For this reason is difficult to find a diurnal pattern in our VOC time
series at this site (see Figures 3e and 3f). In general, concentrations of benzene, toluene, C2-
benzenes and C3-benzenes were lower than 1 ppbv during the entire diurnal cycle. However, a
32
number of spikes with higher concentrations were observed, possibly due to local sources, such
as burning of agricultural debris.
The FOS provided an alternative method to measure concentrations of olefins at real time at
the CENICA site during the MCMA-2003 study. As shown in Figure 4, the diurnal average
pattern of olefinic VOCs detected by the FOS exhibits a similar pattern to what would be
expected for typical pollutants emitted by mobile sources (INE, 2000). The highest olefin
concentrations were measured at 7:00 h and ranged from 30 to 87 ppbv, with 58 ppbv as average.
This morning peak is attributed to the release of anthropogenic emissions into a shallow mixed
layer as the work day. Low olefin concentrations were observed during the afternoon, with an
average of 6.6 ppbv, when dilution through the deep mixed layer was large and emissions were
reduced compared to morning periods. The diurnal olefin pattern was relatively constant during
the entire study. The MCMA-2003 field campaign included the Holy Week (April 14 – 20), a
period in which the vehicular traffic is reduced as many of the city residents leave for the holiday
period. By taking measurements before, during and after this period, we expected to obtain data
to help determine the influence of vehicular emissions upon atmospheric pollution. The FOS
reported a small difference of 6 ppbv in the olefin mixing ratio during the early morning peak
between Holy Week and the other two measurement weeks.
A comprehensive analysis of the diurnal patterns of benzene and toluene measured at La
Merced by DOAS is provided by Grutter and Flores (2004), while Jobson et al. (2005) describe
in detail the diurnal course of a number of VOCs measured by PTR-MS at the CENICA site.
2.6.2. Ambient mixing ratios and hydrocarbon reactivity
We have described that the highest ambient mixing ratios of VOCs in the atmosphere of
Mexico City occur during the morning rush hours (6:00 to 9:00 h). An analysis of the canister
sampling from the four urban monitored sites (Pedregal, La Merced, CENICA and
Constituyentes) during this morning period will provide a description of the VOC species that
contribute to photochemical ozone and haze production in Mexico City. On the basis of average
33
concentrations, the 10 most abundant VOCs for those sites were in decreasing order: propane
(113 ppbv), n-butane (43), ethylene (19), i-butane (16), i-pentane (16), ethane (15), toluene (13),
acetylene (12), n-pentane (7), and MTBE (6). In general, VOC mixing ratios observed in this
study are slightly lower than those reported in previous studies (Arriaga et al., 1997; Mugica et
al., 2002), which is consistent with Arriaga et al. (2004) conclusions that ambient VOC
concentrations have reach a stabilization level or have decreased. The elevated levels of low
molecular weight alkanes measured here are consistent with those reported in the first VOC
study in Mexico City (Blake and Sherwood, 1995), and they are attributable mainly to the
widespread use of LPG, as discussed previously.
It is useful to examine the VOC distribution in an urban area in terms of reactivity with the
hydroxyl radical (OH), which in fact represents the contribution of each VOC specie to the OH
loss rate. Such an OH loss rate is a measure of the initial peroxy radical formation rate, which is
frequently the rate-limiting step in ultimate ozone formation. The actual amount of ozone
produced by a given hydrocarbon depends on their particular oxidation mechanism, the
abundance of other hydrocarbons, and NOx concentrations (Carter, 1994). While realizing that
this approach does not account for the full atmospheric chemistry of the compounds considered,
it provides a useful approximation of their relative contributions to daytime photochemistry. For
this purpose, Table 3 lists the major hydrocarbons in the atmosphere of the Valley of Mexico by
reactivity with OH along with their average ambient concentrations during the morning rush
hour. Average reactivity levels and concentrations are shown in columns according to the site
type: urban, rural (Santa Ana Tlacontenco, La Reforma and Teotihuacan) and industrial
(Xalostoc). For urban and rural sites, the reactivity values were calculated independently for
each monitored site but, as the results were quite similar, only the reactivity values calculated for
the average concentrations of each type of site are presented. VOCs are sorted in descendent
order according to their reactivity in urban sites. The OH reaction rate coefficients shown in
Table 3 correspond to the coefficients published by Atkinson (1994, 1997) at 298 K and 1 atm.
34
Where no information was available, the OH reaction rate coefficient was estimated from
information from similar compounds.
The ten most important hydrocarbons in the urban atmosphere of the Valley of Mexico in
terms of ozone production include two aromatics, five olefins and three alkanes: ethylene,
propylene, propane, n-butane, m,p-xylenes, i-butene, toluene, 2-methyl-1-butene, 2-methyl-2-
butene and i-pentane. It is important to point out that the two most reactive VOCs at urban sites
were also the top two VOCs reported for the industrial site. These species were olefins mainly
emitted by vehicles with high concentrations and high reaction rate coefficients. Also, it is
important to highlight that the elevated concentrations of propane and n-butane are sufficient to
rank these two alkanes among the top 5 VOCs, even though their reactivity rate coefficients are
small compared to those for olefins and aromatics.
Overall, major VOCs in terms of ozone production in the Valley of Mexico correspond to
VOCs of anthropogenic origin. Biogenic VOCs seem to be relatively insignificant. Isoprene
concentrations were low (0.36 ppbv in urban sites) and were assumed to have their origin more
in vehicle exhaust than in vegetation. Vegetation was sparse at the three monitored rural sites,
and therefore isoprene concentrations were also low (0.05 ppbv). At rural sites, low molecular
weight alkanes were the most abundant species, having their origin in LPG leakages as it has
been discussed before. At rural sites, the most reactive hydrocarbon was styrene, with ambient
concentrations similar to those observed at urban sites (0.51 ppbv). The presence of styrene in
these rural environments is attributed to local biomass and trash burning.
A comparison of the reactivity levels between the urban and industrial sites indicates that on
average the industrial location is 1.8 times more reactive than an urban atmosphere. Half of the
measured VOCs at the industrial site had reactivity levels between 1.4 and 2.1 times higher than
at the urban sites. A similar comparison between urban and rural sites shows that an urban
atmosphere is 5.9 times more reactive than a rural atmosphere in the Valley of Mexico.
35
2.6.3. Distribution of VOCs by groups
Figure 5 shows the VOC distributions by groups during the morning (6:00 to 9:00 h) and
afternoon (12:00 to 15:00 h). Note that concentrations are in ppbC. Hereafter all contribution
fractions are based on ppbC, as well. The data shown in Figure 5 were derived from canister
samples collected and analyzed by the WSU and IMP groups. In the morning, the entire valley
experiences a relatively homogeneous mix of VOCs consisting of ~60 % alkanes, ~15 %
aromatics, ~5 % olefins and a remaining 20% of alkynes, halogenated hydrocarbons, oxygenated
species (esters, ethers, etc.) and other unidentified VOCs. In our study, concentrations of total
VOCs in the industrial area were 1.6 times higher than at urban sites during the morning period,
while concentrations at rural sites were about one-sixth of those measured at the urban locations.
In the afternoon, distributions were different with a higher contribution of unidentified VOCs at
the urban sites. The total VOC concentration was 2.4 times lower in the afternoon compared to
the morning at urban sites. However, not all VOC groups showed the same reduction, alkanes
were reduced 70%, olefins 60%, aromatics 53%, and unidentified VOCs 20%. Lower afternoon
concentrations are normally ascribed to increased dispersion and photochemical oxidation during
the midday period. However, employing these mechanisms alone is insufficient because the
relatively unreactive alkane group showed the largest proportional loss. Emission rates must be
a factor, as well, with a large decrease in alkane emissions relative to aromatics and olefins.
Santa Ana Tlacotenco, a rural downwind boundary site, showed an opposite pattern. Transport
from the urban region resulted in afternoon VOC concentrations 2.4 higher than in the morning.
Hydrocarbons. At urban and industrial sites during the morning period, hydrocarbons with
four or less carbons represent the major fraction of the alkanes and alkenes. The main
contributors are propane, n-butane, ethylene, propylene, and the sum of i-, t-2- and -c-2 butenes.
1,3-butadiene is a four-carbon diene that is considered to be a carcinogenic and reproductive
toxicant to humans, whose main source is vehicle exhaust (USEPA, 1999). 1,3-Butadiene is of
concern in Mexico City because of its relatively elevated concentrations: 0.48 and 0.63 ppbv at
urban and industrial sites, respectively.
36
The most abundant aromatics were the BTEX species (benzene, toluene ethylbenzene and
the xylene isomers). They average about 75% of the aromatic burden. Toluene showed an
average concentration of 13.1 ppbv at urban sites and concentrations up to about 33 ppbv in the
industrial area. Total xylene concentrations averaged 14.5 ppbv at the industrial site and about
half that (7. 4 ppbv) in the urban area. Ethylbenzene had average concentrations about the same
as the individual xylene isomers with urban concentrations 1.5 ppbv in the urban area and about
3 ppbv in the industrial region. Benzene averaged 5.1 and 2.9 ppbv at the industrial and urban
sites. The urban benzene concentration determined in this study is similar to that reported by
Bravo et al. (2002) for residential areas in southwest Mexico City.
Oxygenated hydrocarbons. Methyl tertiary-butyl ether (MTBE) is the only oxygenated VOC
listed in Table 3. MTBE is used as an additive in unleaded gasoline to enhance the combustion
efficiency. MTBE contributes no more than 2 % to the total VOC burden of the Valley of
Mexico. Ethyl tertiary butyl ether (ETBE) was another oxygenated VOC identified in the
samples analyzed by IMP. ETBE is not included in Table 3 because it was measured at only half
of the sites (Xalostoc, Pedregal, CENICA and La Merced). Average ETBE concentrations were
0.3 and 0.6 ppbv at urban and industrial sites, respectively. Both, MTBE and ETBE have their
origin in vehicle exhaust due to incomplete combustion and gasoline leakage from fueling
stations.
Halogenated hydrocarbons. The GC-FID technique can be used to identify halogenated
VOCs, but not to quantify precisely their concentration, since halogenated VOCs contain other
atoms besides carbon and hydrogen. However, IMP quantified approximately the ambient
concentrations of 14 halogenated VOCs in ppbC at the Xalostoc, Pedregal, CENICA and La
Merced sites. Table 4 shows that the halogenated species contribute less than 2.5% to the total
VOC burden. Many of them are classified as toxic and carcinogenic species, and others, such as
the vinyl chloride, are suspected to cause congenital malformation. Another concern is the very
long atmospheric residence times of chlorofluorocarbons such as Freon 113. They can
eventually diffuse into the stratosphere where photolysis produces chorine radicals, which
37
catalytically destroy ozone and indirectly contribute to the greenhouse effect. All halocarbons
listed in Table 4 are emitted from anthropogenic sources in urban atmospheres. Even though the
use of chlorofluorocarbons as refrigerants is not allowed anymore, most release of
chlorofluorocarbons occurs during the disposal of refrigeration units and, in developing cities
such as Mexico City, leakage from old residential refrigerators may be a significant source, also.
This may be the reason why ambient concentrations of Freon-113 were higher for urban sites
than for the industrial site.
2.6.4. Comparison of ambient VOC concentrations to vehicles exhaust signatures
Since roadway vehicle emissions are normally the dominant VOC source in urban areas
(Watson et al., 2001), it was of interest to compare the mixing ratios obtained from canister
sampling at the urban sites to the vehicle exhaust source signature measured during mobile
vehicle chase experiments.
In brief, vehicle chase measurements were made using the Aerodyne mobile laboratory
equipped with several instruments to characterize emissions from vehicles under actual driving
conditions (see Herndon et al., 2005). In chase mode, the Aerodyne van was driven immediately
behind a selected vehicle for approximately 10 to 20 minutes. Vehicle plume samples were
collected with an auto-sampling system that included a fast response CO2 sensor (LICOR LI-
7000). Distinct peaks in the CO2 mixing ratio during a vehicle chase indicated interception of
the exhaust plume. When the CO2 levels were elevated above a selected threshold, a conditional
VOC sampler was activated to sample into a “chase” canister and when CO2 levels were below
the threshold, air was channeled into an “urban background” canister.
For the comparison between ambient and vehicular emission data, it is necessary to remove
the impact of photochemical aging on source signatures. This is achieved by regressions
between species with similar atmospheric lifetimes. The source ratio is preserved for species
with similar lifetimes because photochemical loss and mixing will result in similar rates of
concentration change (Parrish et al., 1998). This procedure allowed all of the ambient data to be
38
used including afternoon data when mixing ratios were typically lower due to mixing and
photochemical removal. The slope obtained in ambient data plots defines the source ratio that
can be directly compared to the vehicular chase measurements and literature values. In practice,
there are a limited number of species that can be employed in this analysis because
photochemical loss rates must be similar. We have constrained the hydrocarbon pairings such
that OH rate coefficients differ by 20% or less. The exception is the regression between propene
and ethene where rate constants differ by a factor of three.
Correlations between selected alkenes, alkanes and aromatics are shown in Figures 6, 7 and
8, respectively. The average ratios from the vehicular chase data and from the ambient sampling
at urban, rural and industrial sites are tabulated in Table 5. In addition, Table 5 shows average
exhaust ratios for Mexican gasoline and diesel vehicles from a tunnel study conducted in Mexico
City (Mugica et al., 2001). An average vehicle exhaust ratio for light duty vehicles calculated by
Jobson et al. (2004) from six published tunnel studies conducted in the 1990s in US and Canada
is included in the table, as well.
In the alkene group, t-2-pentene versus c-2-pentene exhibits excellent agreement between
the ambient and vehicular emission ratios. The ambient concentrations span three orders of
magnitude due to atmospheric processing and variations in source strength. In general, the
highest concentrations were recorded during vehicular chase experiments. In a few cases,
ambient industrial and urban samples approached vehicle chase concentrations. The very good
agreement between the exhaust emission and ambient ratios for the 2-pentenes clearly implies a
vehicle exhaust source signature. The t-2-butene versus the c-2-butene correlation shows
reasonable agreement between vehicular chase emissions, ambient ratios and literature values.
The similarity of the ambient and vehicle exhaust ratios for these species suggests vehicle
exhaust as their primary source. For 1-hexene and 1-pentene, ambient data showed considerable
scatter suggesting that each site is impacted by a mix of different sources, and that sources and
emission rates of 1-hexene are not strongly correlated with sources and emission rates of 1-
pentene. The propylene:ethylene ratio displays a fair correlation but contains considerable
39
scatter. The vehicle emission ratio (0.58±0.37) bisects the ambient data, but the large amount of
scatter about this ratio is indicative of multiple independent sources. The emissions ratio itself
has a high degree of scatter, suggesting that the emission relation of these species may vary
considerably within the Mexican fleet. As indicated previously, propylene and ethylene rate
constants vary much more than for the other alkene pairs, which may contribute also to the poor
correlations observed for these species.
Alkane ratios for i-butane:n-butane showed excellent agreement between the ambient,
vehicular emission and literature values. Note that the average ratios in Table 5 for ambient
(0.38 industrial and 0.37 urban) and vehicle exhaust (0.36) are essentially equal. Mugica et al.
(2001) reported exhaust emission ratios of 0.32 and 0.48 for gasoline and diesel vehicles in
Mexico. These results suggest that vehicle exhaust is an important source of n-butane and i-
butane, besides residential LPG usage. Through a source apportionment analysis, Mugica et al.
(2002) determined that vehicle exhaust contributes ~20% to the presence of these two alkanes,
while handling and distribution of LPG contributes ~65%. Although LPG powered vehicles
represent less than 1% of the total fleet, they should be considered also as important sources.
Schifter et al. (2000) evaluated the LPG vehicles program implemented in Mexico City and
found that 95% of them exceed the required environmental regulations. The LPG fleet is
composed mainly by vehicles of intensive use (light- and heavy duty trucks, and small buses
with a 20 passenger capacity). Gasca et al. reported that tailpipe and evaporative emissions of i-
butane contribute 16 and 28%, respectively to the total VOC emissions from LPG vehicles, and
17 and 21% of n-butane.
Good agreement between 2-methylpentane:3-methylpentane ambient, vehicle chase and
literature ratios imply vehicle exhaust emissions as the primary source of these VOCs in the
Mexico City atmosphere. For the other alkane pairs listed in Table 5, agreement between
ambient and vehicle exhaust ratios was not as good. The ambient ratios generally followed the
vehicular emissions line, but with considerable scatter. This suggests that species such as
40
isopentane, n-pentane, hexane, cyclohexane, n-heptane, n-nonane and n-octane are emitted by
vehicles, but also by a variety of other anthropogenic sources.
As illustrated in Figure 8, the ratios of i-propylbenzene and styrene with other aromatics
displayed significant scatter suggesting the importance of other anthropogenic sources, such as
industries and trash burning, for these two VOCs. The xylenes showed excellent agreement with
vehicle exhaust ratios, in fact they showed the best agreement between all correlated VOCs,
indicating clearly that their source is vehicle exhaust. The ethylbenzene:toluene average ratios
from vehicle exhaust and ambient measurements agreed quite well although there was more
scatter in the data. The toluene versus benzene ratio violates the assumption of similar OH
reactivity, but is shown to illustrate the variability. It varied from 3.7 at rural sites to 9.2 at the
industrial site. The median ratio at urban sites (4.9) was similar to the ratio published by Bravo
et al. (2002) for residential areas of Mexico City.
Another method of comparing the ambient data to the roadway vehicle exhaust signature is
to compare ratios using a tracer species as a reference. In this case, we compared hydrocarbon
abundances relative to acetylene since acetylene is known to be a good marker for vehicular fuel
combustion (Barletta et al., 2002). The use of ratios compresses the large differences in
concentration that exist between the various environments and provides a direct comparison of
hydrocarbon distribution patterns. To eliminate changes in the ratio due to chemical aging, the
ambient data were restricted to the morning period between 6:00 and 9:00 h. Figure 9 shows the
median values of the VOC/acetylene ratios together with the 10th and 90th percentile ranges. A
good correlation can be seen between vehicle exhaust and ambient ratios at urban sites with all
exhaust ratios falling within the 10-90 percentile confidence interval. Species with the highest
OH reactivity showed bigger deviations from the exhaust ratio suggesting that, even in samples
collected during the early morning hours, a chemical aging bias may affect the data. Reactive
species such as 1,3-butadiene, isoprene, 1-pentene, 1,2,4-trimethylbenzene and 1,3,5-
trimethylbenzene had median ratios that were between 30 and 45% of the exhaust values. For
example, while the abundances of 1,3-butadiene and 1,3,5-trimethylbenzene relative to acetylene
41
at urban sites were 31 and 44% those of the chasing samples, the ambient MTBE and toluene
ratios with acetylene agreed well with the exhaust ratio. This would support a chemical age
argument because 1,3-butadiene and 1,3,5-trimethylbenzene are about ten times more reactive
with OH than MTBE and toluene. It is important to highlight that the C2-C4 alkanes (ethane,
propane, i-butane and n-butane), and ethyl acetate showed higher ratios compared to exhaust
values at urban sites. This is what would be expected if other anthropogenic sources contribute to
low molecular weight alkane emissions. All olefin and aromatic ratios were lower than the
vehicle chase ratios, which is consistent with an ageing bias. Interesting, the 90th percentile
ambient ratio boundary agrees very well with the vehicle chase ratio. The 90th percentile data
perhaps reflects fresher emissions and less aging bias.
2.6.5. Comparison of ambient VOC concentrations to the emissions inventory
The ambient VOC data can be compared to the distribution of VOC classes represented in
the most recent emissions inventory derived for air quality modeling in Mexico City (West et al.,
2004). This emissions inventory was based on annual emissions reported in 1998 (CAM, 2001).
The official inventory was created by local government authorities using bottom-up methods and
emissions factors which were either measured locally or taken from elsewhere. The VOC
speciation was based on the SAPRC-99 chemical mechanism for VOC reactivity assessment
(Carter, 2000) and a standard mixture of hydrocarbons in urban atmospheres in the United States
(Jeffries et al., 1989). The speciation was determined for each source category using emissions
profiles measured in Mexico City (Mugica et al., 1998, 2002; Vega et al., 2000). These profiles
were adjusted to include species and source categories not measured in Mexico City using
emissions profiles from the SPECIATE database (EPA, 1993).
The median ambient data are lumped into the inventory modeling classes in Table 6. For
comparison, the total emissions by each class are also included in the table along with the
corresponding percentage of the total. The comparison was limited to the morning period
between 6:00 and 9:00 h when concentrations are strongly related to anthropogenic emissions
42
before the photochemistry’s beginning. The column showing the adjustment factor, which takes
into account the molecular weight of each class, reflects the degree of change needed to yield the
same distribution in the emissions inventory as observed in ambient VOC concentrations. This
comparison of early morning ambient data and gridded total emissions suggests that some, but
not all, classes are underestimated in the inventory by factors of 1.1 to 3.1, in contrast to other
species that are overestimated, such as some aromatic and olefin classes (ARO1, ARO2, OLE1
and OLE2). The extreme adjustment factors of 10.8 and 0.05 for butadiene and isoprene,
respectively, are due to their small ambient concentrations compared to other VOC classes.
Overall, the emissions inventory appears to underestimates the contribution of alkanes and
overestimates the contributions of olefins and aromatics. These results do not support the idea
that all VOC emissions reported in the official inventory are underestimated by a factor of 3
(Arriaga et al., 2004; West et al., 2004), however this is a relatively simplistic comparison that
does not fully account for the spatial and temporal distribution of emissions, the small number of
monitoring sites, or for any early morning chemistry that might affect the ambient levels.
2.6.6. Comparison of olefin concentrations measured by FOS to olefin concentrations
calculated by the CIT model
The high resolution and continuous data from instruments such as the FOS provide a basis
for comparison with grid model simulations of selected VOCs. For example, results are shown
in Figure 10 for olefin mixing ratios measured at the CENICA site during the MCMA-2003 field
campaign by FOS and modeled with the California Institute of Technology (CIT) airshed model.
This model was used by West et al. (2004) to model ozone photochemistry in Mexico City.
Concentrations of ethene, isoprene, and groups OLE1 and OLE2 in the CIT model were
weighted according to their contribution to the hydrocarbon mix used by the SAPRC99 chemical
mechanism and their corresponding FOS response factor.
43
The modeled concentrations closely follow the measured concentrations. This result was
unexpected because of all the assumptions and uncertainties involved in the model and the FOS
response. An evaluation of the fractional error defined by:
( )( )( )%1001
1 21∑
= −
−=
N
i om
om
ii
ii
cccc
NErrorFractional (1)
where cm and co represents the modeled and measured concentrations, respectively, showed a
median error of 51% for the entire MCMA-2003 field campaign. During the daytime (6:00 -
19:00 h) the fractional error was smaller than at night (19:00 – 6:00 h), 48% versus 56%. Figure
10a shows there were periods when modeled and measured levels matched very well (day 105),
and other periods when the FOS reported higher or lower concentrations than the model. In
general, the model tends to under-predict concentrations at night and to over-predict during the
day. Figure 10b shows that the major discrepancy occurs during the rush hour when the model
over-predicts the morning peak. This peak is predicted by the model to occur one hour later than
observed in the ambient measurements. Overall, the estimated olefin concentrations are in good
agreement with the measured concentrations, since modeled concentrations are in the one
standard deviation range of the measured concentrations.
This result is consistent with our conclusions drawn from the comparison with the emissions
reported in the emissions inventory, and from eddy covariance flux measurements of selected
VOCs performed during the MCMA-2003 (Velasco et al., 2005). These analyses suggest that
emissions of some VOC classes reported in the official emissions inventory are essentially
correct, while others are underestimated by factors between 1.1 and 3. If this is correct, previous
suggestions that poor agreement between modeled and measured ozone levels in Mexico City
was due to a threefold underestimation of total VOC emissions (Arriaga-Colina et al., 2004;
West et al., 2004) may not be correct. Quite possibly other factors, such as incorrect NOx
emissions or an erroneous evolution of the boundary layer may be causing the poor ozone
agreement.
44
2.7. Summary and conclusions
A number of independent methods were used to measure ambient VOC concentrations in the
Valley of Mexico during the MCMA 2002 and 2003 field campaigns. The use of different
techniques allowed a wide range of individual species to be measured with different spatial and
temporal scales, providing confidence in the data, as well as a basis for comparison with grid
model simulations of selected VOCs. The VOC concentrations were analyzed to understand
better their distribution, diurnal pattern, origin and reactivity in the atmosphere of Mexico City.
All data have been valuable to evaluate the Mexican emissions inventory, as well to support
effective control strategies. The following points summarize the main findings and conclusions
of this work.
- Ambient air inter-comparison of GC, PTR-MS and DOAS measurements for selected VOCs
reported good agreement. This allows to combine data into a single comprehensive data set
including measurements of VOCs concentrations of different spatial and temporal scales.
- At urban sites, the ambient concentration of VOCs depends strongly on the vehicular activity.
At La Merced, no late afternoon peak in VOC concentration was observed, while at the Pedregal
site the morning and late afternoon VOC peaks were observed. At both sites, the VOC peaks
corresponded to the vehicular traffic peaks during the morning and afternoon rush hours.
- At boundary sites, the diurnal profiles of ambient VOC concentrations depend mainly on wind
patterns. These sites correspond to rural landscapes, where burning of agriculture debris and
trash is common. Biomass burning was observed to produce spikes on ambient concentrations of
selected VOCs, such as styrene.
- In general, VOC concentrations reported here are smaller than concentrations reported in the
last decade. This finding is consistent with Arriaga et al. (2004), who state that ambient VOC
concentrations have stabilized and possibly decreased. This is a good indicator that enacted
policies and actions to control VOC emissions have shown success, despite the growth in the
vehicular fleet and other activities.
45
- In the morning, the entire valley experiences a relatively homogeneous mix of VOCs consisting
of ~60 % alkanes, ~15 % aromatics, ~5 % olefins and a remaining 20% of alkynes, halogenated
hydrocarbons, oxygenated species and other unidentified VOCs. Contributions are based on
ppbC. In the afternoon, distributions were different with a higher contribution of unidentified
VOCs at the urban sites.
- In terms of ozone production, olefins are the hydrocarbons of major concern in Mexico City;
ethylene and propylene are the two most reactive VOCs. The elevated concentrations of propane
and n-butane are sufficient to rank these two alkanes among the top 5 VOCs, even though their
reactivity rate coefficients are small compared to those for olefins and aromatics.
- Elevated levels of toxic VOCs, such as 1,3-butadiene, vinyl chloride and the BTEX
hydrocarbons were observed. These VOCs are of public health concern.
- The ratios between ambient concentrations of two hydrocarbons of similar lifetime and the
ratios of each different VOC with acetylene demonstrated that many olefins and aromatics have
their main origin in vehicle exhaust, such as the xylenes, toluene, ethylbenzene, t-2-pentene, c-2-
pentene, etc., as well as MTBE and some alkanes, such as 2-methylpentane and 3-
methylpentane. Besides, the ratios with acetylene showed that vehicle exhaust contributes to the
emission of all VOC species, including alkanes of light molecular weight that have been widely
related with the use of residential LPG, particularly propane and n-butane.
- Examination of the VOC data in terms of the relative distribution of lumped modeling VOC
classes and comparison to the emissions inventory suggests that the inventory underestimates the
contribution of some alkanes and overestimates the contributions of some olefins and aromatics.
- The comparison between ambient concentrations of olefins measured by FOS and olefins
predicted by the CIT model at the CENICA site showed relatively good agreement. This finding
does not support the statement that the emissions inventory underestimates the emissions of
VOCs by a factor of 3.
Although some pollutants have been successfully controlled in Mexico City during the last
decade, other pollutants are still reporting elevated concentrations. Among these pollutants are
46
several VOCs discussed in this manuscript. Effective control strategies to reduce ambient VOC
concentrations need to consider both, the VOC reactivity in terms of production of ozone and
other secondary pollutants, and the toxic potential. These control strategies must be focused in
improve fuels quality and vehicle technology, since there are strong evidence that vehicle
exhaust is the main source of many hydrocarbons. However, they are not going to solve by
themselves the air pollution problem in Mexico City, where not all population have access to
efficient transport and clean technology for basic daily activities, such as cooking and water
heating. The economic necessities of the major population need to be solved, and the actual style
of life needs to be modified, as well. Private cars cannot continue being the main transport
mode, new types of massive public transport with excellent quality need to be implemented to
promote the less use of private cars. A good example is the system of confined buses to unique
lanes in main avenues simulating subways lines above the surface, such as the “Metrobus
system” in Mexico City, which requires a relatively inexpensive infrastructure compared to other
systems. Also, new policies promoting environmental education, the use of bicycle and a greater
support to environmental research are needed to reduce the harmful VOC levels reported here,
and overall, the air pollution problem in Mexico City.
2.8. References
Arriaga, J.L., Escalona, S., Cervantes, A.D., Orduñez, R. & Lopez, T. Seguimiento de COV en aire urbano de la ZMCM 1992-1996. In Contaminación Atmosférica, vol. 2, edited by L.G. Colin and J.R. Varela, pp 67-98, UAM, Mexico D.F., 1997.
Arriaga-Colina, J.L., West, J.J., Sosa, G., Escalona, S.S., Orduñez, R.M. & Cervantes, A.D.M. Measurements of VOCs in Mexico City (1992-2001) and evaluation of VOCs and CO in the emissions inventory. Atmospheric Environment 38, 2523-2533 (2004).
Atkinson, R. Gas-phase tropospheric chemistry of organic compounds. Journal of Physical and Chemical Reference Data Monographs 2, 11-216 (1994).
Atkinson, R. Gas-phase tropospheric chemistry of volatile organic compounds. Journal of Physical and Chemical Reference Data Monographs 26, 215-290 (1997).
47
Barletta, B., Meinardi, S., Simpson, I.J., Khwaja, H.A., Blake, D.R. & Rowland, F.S. Mixing ratios of volatile organic compounds (VOCs) in the atmosphere of Karachi, Pakistan. Atmospheric Environment 36, 3429-3443 (2002).
Blake, D.R. & Sherwood, F. Urban leakage of liquefied petroleum gases and its impact on Mexico City air quality. Science 269, 953-956 (1995).
Bravo, H., Sosa, R., Sanchez, P., Bueno, E. & Gonzalez, L. Concentration of benzene and toluene of the southwestern area at the Mexico City Metropolitan Zone. Atmospheric Environment 36, 3843-3849 (2002).
Carter, W. P. L. Development of ozone reactivity scales for volatile organic compounds, Journal of Air & Waste Management Association 44, 881– 899 (1994).
Carter, W.P.L. Documentation of the SAPRC-99 Chemical Mechanism for VOC reactivity assessment, final report to California Air Resources Board, contracts 92-329 and 95-308. (Calif. Air Res. Board, Sacramento, California 2000, available at www.cert.ucr.edu/~carter).
Comisión Ambiental Metropolitana (CAM). Inventario de emisiones a la atmósfera, Zona Metropolitana del Valle de México, 1998. (Comisión Ambiental Metropolitana, México, D.F. 2001).
Conner, T.L., Lonneman, W.A. & Seila, R.L. Transportation related volatile hydrocarbon source profiles measured in Atlanta. Journal of Air & Waste Management Association 45, 383–394 (1995).
Cruz-Nuñez, X., Hernandez-Solis, J.M. & Ruiz-Suarez, L.G. Evaluation of vapor recovery systems efficiency and personal exposure in service stations in Mexico City. The Science of the Total Environment 309, 59-68 (2003).
de Gouw, J.A., Goldan, P.A., Warnake, C., Kuster, W.C., Roberts, J.M., Marchewka, M., Bertman, S.B., Pszenny, A.A.P. & Keene, W.C. Validation of proton transfer reaction-mass spectrometry (PTR-MS) measurements of gas-phase organic compounds in the atmosphere during the New England Air Quality Study (NEAQS) in 2002. Journal of Geophysical Research 108, 4682-4199 (2003).
Doran, J.C. & Zhong, S. Thermally driven gap winds into the Mexico City basin. Journal of Applied Meteorology 39, 1330-1340 (2000).
Evans, J., et al. in Quality in the Mexico Megacity: an integrated assessment (ed. Molina, L.T. & Molina M.J.) 105-136 (Kluwer Academic Publishers, Netherlands, 2002).
Finlayson-Pitts, B.J. & Pitts, J.N. Chemistry of the upper and lower atmosphere. (Academic Press, USA 2000).
Fraser, M.P., Cass, G.R., & Simoneit, B.R.T. Gas-phase and particle-phase organic compounds emitted from motor vehicle traffic in a Los Angeles roadway tunnel. Environmental Science and Technology 32, 2051–2060 (1998).
48
Gakenheimer, R., Molina, L.T., Sussman, J., Zegras, C., Howitt, A., Makler, J., Lacy, R., Slott, R. & Villegas A. The MCMA transportation system: mobility and pollution. In Air Quality in the Mexico Megacity: An Integrated Assessment., edited by Molina, L.T. & Molina M.J., pp 213-284. Kluwer Academic Publishers, Netherlands, 2002.
Gasca, J., Ortiz, E., Castillo, H., Jaimes, J.L. & Gonzalez, U. The impact of liquefied petroleum gas usage on air Quality in Mexico City. Atmospheric Environment 38, 3517-3527 (2004).
Gobierno del Distrito Federal (GDF). Informe del Estado de la Calidad del Aire y Tendencias 2003 para la Zona Metropolitana del Valle de México. (Dirección General de Gestión Ambiental del Aire. México, D.F., 2004).
Gomez-Perales, J.E., et al. Commuters’ exposure to PM2.5, CO, and benzene in public transport in the metropolitan area of Mexico City. Atmospheric Environment 38, 1219-1229 (2004).
Grutter, M. & Flores, E. Air pollution monitoring with two optical remote sensing techniques in Mexico City. Proceedings of SPIE 5571, Gran Canaria, Spain, 13-16 of September (2004).
Guenther, A. & Hills, A. Eddy covariance measurement of isoprene fluxes. Journal of Geophysical Research 103, 13145-13152 (1998).
Herndon, S.C., Kolb, C.E., Lamb, B., Westberg, H., Allwine, E., Velasco, E., Knighton, B., Zavala, M., Molina, L. & Molina, M. Conditional sampling of volatile organic compounds in on-road vehicle plumes. In preparation (2005).
Instituto Nacional de Ecología (INE). Almanaque de datos y tendencias de la calidad del aire en ciudades mexicanas. (INE-SEMARNAT. México, D.F., 2000).
Jazcilevich, A.D., Garcia, A.R. & Ruiz-Suarez, L.G. A study of air flow patterns affecting pollutant concentrations in the Central Region of Mexico. Atmospheric Environment 37,183-193 (2003).
Jeffries, H.E., Sexton, K.G., Arnold, J.R. & Kale T.L. Validation testing of new mechanisms with outdoor chamber data, Volume 2: analysis of VOC data for the CBA and CAL photochemical mechanisms, final report EPA-600/3-89-010b. (Environmental Protection Agency, Research Triangle Park, N.C. 1989).
Jobson, B.T., Alexander, M.L., Prazeller, P., Berkowitz, C.M., Westberg, H., Velasco, E., Allwine, E., Lamb, B., Volkamer, R., Molina, L. & Molina, M. Intercomparison of volatile organic carbon measurements techniques and data from the MCMA 2003 field experiment. In preparation (2005).
Jobson, B.T., Berkowitz, C.M., Kuster, W.C., Goldan, P.D., Williams, E.J., Fesenfeld, F.C., Apel, E.C., Karl, T., Lonneman, W.A. & Riemer, D. Hydrocarbon source signatures in Houston, Texas: influence of the photochemical industry. Journal of Geophysical Research 109, D24305,doi10.1029/2004JD004887 (2004).
49
Kirchstetter, T.W., Singer, B.C., Harley, R.A., Kendall, G.R. & Chan W. Impact of oxygenated gasoline use on California light duty vehicle emissions, Environmental Science and Technology 30, 661– 670 (1996).
Lindinger, W., Hansel, A. & Jordan, A. On-line monitoring of volatile organic compounds at pptv levels by means of proton-transfer-reaction mass spectroscopy (PTR-MS): Medical applications, food control and environmental research. International Journal of Mass Spectrometry Ion Processes 173, 191-241 (1998).
Meneses, F., Romieu, I., Ramirez, M., Colome, S., Kochy, F., Ashley, D. & Hernandez-Avila, M. A survey of personal exposures to benzene in Mexico city. Archives of Environmental Health 54, 359-363 (1999).
Molina, M.J. & Molina, L.T. Megacities and atmospheric pollution. Air and Waste Management Association 54, 644-680 (2004).
Mugica, V., Ruiz, M.E., Watson, J. and Chow, J. Volatile aromatic compounds in Mexico City atmosphere: levels and source apportionment. Atmósfera 16, 15-27 (2003).
Mugica, V., Vega, E., Arriaga, J.L. & Ruiz, M.E. Determination of motor vehicle profiles for non-methane organic compounds in the Mexico City Metropolitan Area. Journal of Air & Waste Management Association 48, 1060-1068 (1998).
Mugica, V., Vega, E., Chow, J., Reyes, E., Sanchez, G., Arriaga, J.L., Egami, R. & Watson, J. Speciated non-methane organic compounds emissions from food cooking in Mexico. Atmospheric Environment 35, 1729-1734 (2001).
Mugica, V., Vega, E., Ruiz, H., Sanchez, G., Reyes, E. & Cervantes, A. Photochemical reactivity and sources of individual VOCs in Mexico City. In Air pollution X (ed. Brebbia, C.A. & Martin-Duque, J.F.) WIT Press, UK 2002.
Mugica, V., Watson, J.,Vega, E., Reyes, E., Ruiz, M.E. & Chow, J. Receptor model source apportionment of nonmethane hydrocarbons in Mexico City. The Scientific World Journal (2), 844– 860 (2002).
Ortiz, E., Alemon, E., Romero, D., Arriaga, J.L., Olaya, P., Guzman, F. & Rios, C. Personal exposure to benzene, toluene and xylene in different microenvironments at the Mexico City metropolitan zone. The Science of the Total Environment 287, 241-248 (2002).
Parrish, D.D., Trainer, M., Young, V., Goldan, P.D., Kuster, W.C., Jobson, B.T., Fehsenfeld, F.C., Lonneman, W.A., Zika, R.D., Farmer, C.T., Riemer, D.D. & Rodgers, M.O. Internal consistency tests for evaluation of measurements of anthropogenic hydrocarbons in the troposphere. Journal of Geophysical Research 103, 22339–22359 (1998).
Raga, G.B., Baumgardner, D., Castro, T., Martinez-Arroyo, A. & Navarro-Gonzalez, R. Mexico City air quality: a qualitative review of gas and aerosol measurements (1960-2000). Atmospheric Environment 35, 4041-4058 (2001).
50
Rogak, S.N., Pott, U., Dann, T. & Wang D. Gaseous emissions from vehicles in a traffic tunnel in Vancouver, British Columbia. Journal of Air & Waste Management Association 48, 604– 615 (1998).
Sagebiel, J.C., Zielinska, B., Pierson, W.R. & Gertler, A.W. Real world emissions and calculated reactivities of organic species from motor vehicles. Atmospheric Environment 30, 2287– 2296 (1996).
Schifter, I., Diaz, L., Duran, J., Guzman, E., Chavez, O. & Lopez-Salina, E. Remote sensing study of emissions from motor vehicles in the Metropolitan Area of Mexico City. Environmental Science and Technology 37, 395-401 (2003).
Schifter, I., Diaz, L., Lopez-Salinas, E., Rodriguez, R., Avalos, S. & Guerrero, V. 2000. An evaluation of the LPG vehicles program in the metropolitan area of Mexico City. Journal of the Air and Waste Management Association 50,301–309 (2000).
Shiohara, N., Fernández-Bremauntz, A., Blanco-Jiménez, S. & Yanagisawa Y. The commuters’ exposure to volatile chemicals and carcinogenic risk in Mexico City. Atmospheric Environment 39, 3481-3489 (2005).
US Environmental Protection Agency (USEPA). Compendium method TO-14a. Determination of volatile organic compounds (VOCs) in ambient air using specially prepared canisters with subsequent analysis by gas chromatography (EPA/625/R-96/010b). Compendium of methods for the determination of toxic organic compounds in ambient air, 2nd Edition. US EPA, Cincinnati, OH, 1999.
US Environmental Protection Agency (USEPA). Integrated Risk Information System (IRIS) on 1,3-Butadiene. (National Center for Environmental Assessment, Office of Research and Development, Washington, DC. 1999).
US Environmental Protection Agency (USEPA). SPECIATE, VOC/PM speciation data system, version 1.50. (Environmental Protection Agency and Radian Corporation, Research Triangle Park, N.C. 1993).
Vega, E., Mugica, M., Carmona, R. & Valencia, E. Hydrocarbon source apportionment in Mexico City using the chemical mass balance receptor model. Atmospheric Environment 34, 4121-4129 (2000).
Velasco, E. Estimates for biogenic non-methane hydrocarbons and nitric oxide emissions in the Valley of Mexico. Atmospheric Environment 37, 625-637 (2003).
Velasco, E., Lamb, B., Pressley, S., Allwine, E., Westberg, H., Jobson, T., Alexander, M., Prazeller, P., Molina, L. & Molina, M. Flux measurements of volatile organic compounds from an urban landscape. Geophysical Research Letters 32(20) L20802, (2005).
Warneke, C., de Gouw, J.A., Kuster, W.C., Goldan, P.D. & Fall, R. Validation of atmospheric VOC measurements by proton transfer reaction mass spectroscopy using a gas
51
chromatographic preseparation method. Environmental Science and Technology 37, 2494-2501 (2003).
Watson, J.G., Chow, J.C. & Fujita, E.M. Review of volatile organic compound source apportionment by chemical mass balance. Atmospheric Environment 35, 1567– 1584 (2001).
West, J., Zavala, M.A., Molina, L.T., Molina, M.J., San Martini, F., McRae, J., Sosa, G. & Arriaga-Colina, J.L. Modeling ozone photochemistry and evaluation of hydrocarbon emissions in the Mexico City metropolitan area. Journal of Geophysical Research 109, 19312-19327 (2004).
Whiteman, C.D., Zhong, S., Bian, X., Fast, J.D. & Doran, J.C. Boundary layer evolution and regional-scale diurnal circulations over the Mexico Basin and Mexican plateau. Journal of Geophysical Research 106(D8), 10081-10102 (2000).
52
Table 2.1. Description of the VOC monitoring sites during the MCMA-2002 and 2003 field
campaigns
Site # Site and position
Year Site description Method
1 CENICA (N19.358º, W99.073º)
2003
MCMA-2003 super site located in the Autonomous Metropolitan University campus Iztapalapa at the southeast of the city in a mixed area with residences, light and medium industries, services and commerce. The traffic is heavy and composed by old and new vehicles on paved roads.
Canister sampling and GC-FID analysis. PTR-MS. FTIR & DOAS. FIS.
2 Constituyentes (N19.400º, W99.210º)
2002
Western suburban neighborhood close to Chapultepec park. Heavy traffic on paved roads composed by private cars and heavy-duty diesel buses.
Canister sampling and GC-FID analysis.
3 La Merced (N19.424º, W99.119º)
2002 &
2003
Central city section composed by a mix of residences, small and medium commerce, light industries and a busy market. Heavy traffic on paved roads with private cars, light–duty vehicles and modern heavy-duty diesel buses.
Canister sampling and GC-FID analysis. PTR-MS. FTIR & DOAS.
4 La Reforma (N19.976º, W98.697º)
2003
Southwestern downwind site from Pachuca, city with 245,000 inhabitants located to the northeast of Mexico City. Rural site close to urban areas with reduced vehicular traffic on paved and unpaved roads.
Canister sampling and GC-FID analysis.
5 Pedregal (N19.325º, W99.204)
2002 &
2003
Southwestern suburban neighborhood with paved residential roads lightly traveled. This site is in the prevailing downwind direction from the center of the city.
Canister sampling and GC-FID analysis. PTR-MS.
6
Santa Ana Tlacotenco (N19.177º, W98.99º)
2003
Rural site to the southwest of Mexico City, close to the gap in the mountains at Amecameca. Paved and unpaved roads with minimum traffic.
Canister sampling and GC-FID analysis. PTR-MS.
7 Tehotihucan (N19.688º, W98.870º)
2002
Northern upwind boundary site of Mexico City with pollution influence from a large power plant and large industries around the region.
Canister sampling and GC-FID analysis.
8 Xalostoc (N19.527º, W99.076º)
2002 &
2003
North-eastern industrial section of the city with light to medium industries. Heavy traffic on paved and unpaved roads formed by a mix of new and old gasoline and diesel vehicles.
Canister sampling and GC-FID analysis.
53
Table 2.2. Sensitivities and average concentrations measured during selected days throughout
the MCMA-2003 campaign between 6 and 10 am by a canister sampling system and GC-FID
analysis, FOS responses to those average concentrations, and relative sensitivities to propylene
for six compounds
Compound Sensitivity (photons ppb-1 s-1)
Average conc., 6–10 am (ppb)
Average FOS response (photons s-1) c
Relative sensitivity to propylene d
Propylene 25.4 7.50 191 1.00 Isoprene 74.7 0.304 23 2.94 Ethylene 17.7 20.7 366 0.70 1-3 Butadiene 49.8 0.791 39 1.96 1-Butene 7.9 3.90 a 31 0.31 NO ~0.0 61.2 b 0 ~0.0
a As i-butene b From continuous monitoring during the entire campaign. c Average response = (sensitivity)(average conc.) + (zero value) d Relative sensitivity = (compound sensitivity) / (propylene sensitivity)
54
Table 2.3. OH reactivity and average ambient concentrations of major VOCs measured during the
MCMA-2002 and 2003 studies at urban (Pedregal, La Merced, CENICA and Constituyentes), rural
(Santa Ana Tlacotenco, Teotihuacan and La Reforma) and industrial (Xalostoc) sites of the Valley
of Mexico. Data correspond to the morning rush hours (6:00 to 9:00 h)
OH reactivity§ Ambient concentration (ppbv) Species Group
OH reaction
rate coeff.* Urban Rural Industrial urban rural Industrial Ethylene olefin 8.52 3.91 0.28 6.72 18.7 1.32 32.1
Propylene olefin 26.3 3.43 0.53 6.06 5.30 0.81 9.37
Propane alkane 1.15 3.20 0.46 4.53 113 16.4 160
n-butane alkane 2.54 2.71 0.38 3.81 43.4 6.08 61.4
m,p-xylene aromatic 18.9 2.53 0.26 4.93 5.44 0.56 10.6
i-butene olefin 31.4 2.13 0.46 4.08 2.75 0.60 5.28
Toluene aromatic 5.96 1.93 0.20 4.81 13.2 1.37 32.8
2-methyl-1-butene olefin 61.0 1.67 0.08 3.70 1.11 0.05 2.47
2-methyl-2-butene olefin 86.9 1.57 0.16 2.37 0.74 0.08 1.11
i-pentane alkane 3.70 1.42 0.37 2.61 15.6 4.05 28.7
1,2,4-trimethylbenzene aromatic 32.5 1.41 0.22 2.76 1.76 0.28 3.45
t-2-butene olefin 64.0 1.33 0.42 3.14 0.85 0.27 2.00
t-2-pentene olefin 67.0 1.00 0.15 2.18 0.61 0.09 1.32
c-2-butene olefin 56.4 0.94 0.29 1.73 0.68 0.21 1.25
Isoprene olefin 101 0.88 0.13 0.68 0.36 0.05 0.27
i-butane alkane 2.19 0.87 0.14 1.30 16.1 2.54 24.1
1,3,5-trimethylbenzene aromatic 57.5 0.82 0.17 1.51 0.58 0.12 1.07
1,3-butadiene olefin 66.6 0.79 0.18 1.03 0.48 0.11 0.63
Styrene aromatic 58.0 0.69 0.72 1.63 0.49 0.51 1.14
o-xylene Aromatic 13.7 0.66 0.06 1.31 1.96 0.18 3.88
n-pentane alkane 3.94 0.66 0.06 0.86 6.80 0.65 8.85
2-methylpentane alkane 5.60 0.64 0.06 1.25 4.66 0.47 9.07
Hexane alkane 5.61 0.61 0.03 1.30 4.39 0.23 9.44
c-2-pentene olefin 65.0 0.50 0.27 1.08 0.31 0.17 0.68
MTBE oxygenated 2.90 0.45 0.03 0.59 6.30 0.41 8.32
3-methylpentane alkane 5.70 0.42 0.04 0.88 3.00 0.32 6.30
1-pentene olefin 31.4 0.39 0.19 1.98 0.50 0.24 2.57
3-methylhexane alkane 7.20 0.38 0.07 0.45 2.14 0.41 2.53
2,3-dimethylbutane alkane 5.99 0.35 0.03 0.53 2.40 0.18 3.57
Methylcyclopentane alkane 8.80 0.31 0.04 0.55 1.41 0.18 2.55
m-ethyltoluene aromatic 19.2 0.30 0.03 0.65 0.65 0.06 1.37
Acetylene alkyne 0.91 0.27 0.01 0.35 12.1 0.62 15.7
Ethylbenzene aromatic 7.10 0.27 0.03 0.52 1.54 0.17 2.98
n-heptane alkane 7.15 0.26 0.02 0.38 1.50 0.12 2.13
2-methylhexane alkane 6.80 0.26 0.04 0.37 1.55 0.26 2.20
*OH reaction rate coefficients at 298 K and 1 atm (cm3 molecule-1 s-1) × 10-12, (Atkinson, 1994; Atkinson, 1997). §Products of mean individual VOC concentrations and OH reaction rate coefficients.
55
Continuation table 2.3
OH reactivity§ Ambient concentration (ppbv) Species Group
OH reaction
rate coeff.* Urban Rural Industrial urban rural Industrial i-octane alkane 3.57 0.25 0.02 0.48 2.86 0.23 5.45
ethyl acetate ester 8.00 0.25 0.02 0.00 1.27 0.10 0.00
p-ethyltoluene aromatic 12.1 0.22 0.03 0.32 0.72 0.09 1.08
2,3,4-trimethylpentane alkane 7.00 0.21 0.01 0.31 1.19 0.05 1.82
Cyclohexane alkane 7.49 0.17 0.03 0.37 0.90 0.17 2.02
1-hexene olefin 37.0 0.16 0.15 0.44 0.17 0.17 0.48
2,3-dimethylpentane alkane 7.20 0.16 0.02 0.20 0.89 0.09 1.11
o-ethyltoluene aromatic 12.3 0.15 0.04 0.28 0.50 0.13 0.93
2,4-dimethylhexane alkane 8.60 0.13 0.01 0.16 0.63 0.05 0.73
Methylcyclohexane alkane 10.4 0.13 0.02 0.24 0.49 0.09 0.93
n-octane alkane 8.70 0.12 0.02 0.30 0.56 0.10 1.38
2,2-dimethylbutane alkane 2.59 0.10 0.01 0.17 1.62 0.13 2.61
n-decane alkane 11.2 0.10 0.02 0.23 0.37 0.08 0.85
Ethane alkane 0.27 0.10 0.01 0.28 14.9 1.26 41.8
Nonane alkane 10.0 0.10 0.01 0.24 0.39 0.04 1.00
Benzene aromatic 1.23 0.09 0.02 0.15 2.86 0.53 5.05
2-methylheptane alkane 8.20 0.08 0.02 0.16 0.40 0.10 0.77
Propyne alkyne 5.90 0.08 0.04 0.15 0.52 0.25 1.04
n-propylbenzene aromatic 6.00 0.06 0.01 0.10 0.39 0.05 0.67
2,5-dimethylhexane alkane 8.30 0.05 0.13 0.00 0.26 0.61 0.00
1,2,4 trimethyl cyclohexane alkane 7.50 0.04 0.01 0.08 0.23 0.04 0.42
Cyclopentane alkane 5.02 0.04 0.00 0.02 0.34 0.04 0.13
*OH reaction rate coefficients at 298 K and 1 atm (cm3 molecule-1 s-1) × 10-12, (Atkinson, 1994; Atkinson, 1997). §Products of mean individual VOC concentrations and OH reaction rate coefficients.
56
Table 2.4. Ambient average concentrations of halogenated VOCs measured during the MCMA-
2003 study at urban (Pedregal, La Merced and CENICA) and industrial (Xalostoc) sites of the
Valley of Mexico
Urban sites (ppbC)
Industrial site (ppbC) Species
6-9 am 12-15 h 6-9 h 12-15 h p-dichlorobenzene 5.53 2.26 5.94 2.80 Trichloroethylene 4.15 1.56 4.44 1.48 1,2-dichloropropane 3.51 2.82 0.00 0.00 c-1,3-dichlopropene 3.20 1.67 6.03 3.96 1,2-dichloroethane 3.17 2.27 0.52 0.00 o-dichlorobenzene 2.96 0.39 4.60 0.00 1,1-dichloroethane 1.69 2.01 1.40 2.28 Chloroform 1.36 0.37 3.47 0.93 t-1,3-dichloropropene 0.92 0.26 1.46 0.30 Freon113 0.90 0.94 0.52 0.00 Perchloroethylene 0.51 0.34 0.92 0.28 Chlorobenznene 0.11 0.01 0.20 0.73 Vinyl chloride 0.00 0.35 0.52 0.86 m-dichlorobenzene 0.00 0.67 0.00 0.00 Subtotal halogens 28.01 15.94 30.02 13.62 Halogens percent between unidentified VOCs 11.4 8.2 11.3 3.6
Halogens percent between total VOCs 1.8 2.4 1.2 0.9
57
Table 2.5. Hydrocarbon molar ratios (ppbC/ppbC) measured at the industrial, urban and rural sites and from vehicle chase operation during the MCMA-2002 and 2003 field studies. Also ratios from vehicles exhaust studies in North America are shown for comparison
Industrial Urban Rural Vehicular chases Mexican gasoline vehicles§
Mexican diesel
vehicles§
US & Canada light duty vehicles#
Average* Median Average* Median Average* Median Average* Median Average Average Average*
Alkene ratios t-2-butene/c-2-butene 1.69±0.91 1.25 1.44±0.83 1.21 1.63±0.66 1.48 1.16±0.50 1.10 1.17 1.00 1.27±0.33
t-2-pentene/c-2-pentene 1.82±0.51 1.91 2.05±0.53 2.07 1.83±0.51 1.89 1.98±0.46 1.98 --- 9.33 1.78±0.19
1-hexene/1-pentene 0.63±0.46 0.52 0.67±0.48 0.59 --- --- --- --- --- --- 0.49±0.25
1,3-butadiene/t-2-pentene 1.13±0.58 1.14 0.89±0.52 0.77 0.67±0.47 0.64 1.16±0.97 0.96 --- --- 1.36±0.51
propene/ethane 0.52±0.30 0.44 0.59±0.47 0.42 0.47±0.28 0.44 0.58±0.37 0.50 0.55 0.71 0.30±0.04
Alkane ratios i-butane/n-butane 0.38±0.04 0.37 0.37±0.03 0.37 0.41±0.07 0.41 0.36±0.04 0.36 0.32 0.48 0.19±0.08
i-pentane/n-pentane 3.64±0.69 3.62 3.06±1.41 2.59 9.03±6.61 8.08 2.80±1.69 2.36 2.63 4.27 2.97±0.57
2-methylpentane/3-methylpentane 1.38±0.19 1.41 1.67±1.80 1.54 1.74±0.78 1.59 1.49±0.14 1.47 1.67 1.18 1.69±0.11
Hexane/2-methylpentane 1.15±0.31 1.11 0.80±0.40 0.72 0.90±0.83 0.69 0.65±0.15 0.65 0.78 1.02 0.52±0.05
cyclohexane/n-heptane 1.06±1.03 0.50 0.70±0.60 0.50 1.02±0.74 0.91 0.73±0.63 0.54 0.86 0.29 0.43±0.17
n-octane/nonane 0.87±0.38 1.01 1.56±0.82 1.29 1.26±0.62 1.12 1.36±0.43 1.28 1.04 0.55 1.49±0.05
Aromatic ratios Toluene/benzene 9.17±4.57 8.81 5.42±2.32 4.89 3.74±1.35 3.45 4.14±1.34 3.80 2.78 5.05 1.59±0.28
ethylbenzene/toluene 0.10±0.04 0.09 0.13±0.05 0.13 0.16±0.07 0.15 0.15±0.05 0.15 0.21 0.30 0.17±0.03
Isopropylbenzene/toluene 0.014±0.005 0.014 0.020±0.008 0.020 --- --- --- --- 0.027 0.016 0.020±0.008
o-xylene/m,p-xylene 0.36±0.03 0.36 0.41±0.07 0.39 0.40±0.11 0.40 0.39±0.05 0.39 0.39 0.36 0.38±0.02
Styrene/1,3,5-trimethylbenzene 1.08±0.68 0.98 1.78±1.67 1.22 4.28±3.58 3.10 0.66±0.42 0.55 0.37 0.84 0.81±0.21
1,2,3-trimethylbenzene/1,2,4-trimethylbenzene 0.39±0.09 0.36 0.41±0.13 0.37 --- --- --- --- --- --- 0.25±0.07
* Average ± 1 standard deviation. § From vehicle emission profiles measured in Mexico City in 1998 (Mugica et al., 2001). # Average vehicle exhaust ratio calculated by Jobson et al. (2004) from six tunnel studies conducted in the 1990s (Conner et al., 1995; Kirchstetter et al., 1996; Sagebiel et al., 1996; Fraser et al., 1998; Rogak et al., 1998).
58
Table 2.6. Comparison between ambient VOC concentrations measured at urban sites during the
morning period (6:00 to 9:00 h) and the corresponding VOC emissions from the most recent
emissions inventory for modeling purposes
Model species Ambient [VOC] (ppbC)
% of total
Inventory ×103 (tons/yr)
% of inventory in C moles
Adjustment factor to correct the inventory
ETHANE 29.7 2.3 3.3 1.0 2.3 PROPANE 339.5 25.7 50.1 15.1 1.7 ALK1 470.8 35.7 36.2 11.7 3.1 ALK2 91.5 6.9 78.3 22.5 0.3 ACETYLENE 24.3 1.8 4.2 1.4 1.3 ETHYLENE 37.3 2.8 7.9 2.50 1.1 OLE1 29.5 2.2 15.8 5.2 0.4 OLE2 20.8 1.6 22.9 7.5 0.2 BUTADIENE 1.9 0.2 0.04 0.01 10.8 ISOPRENE 1.8 0.1 7.8 2.5 0.05 BENZENE 17.2 1.3 3.3 1.1 1.1 ARO1 95.5 7.2 40.1 13.0 0.6 ARO2 127.5 9.7 42.8 13.1 0.7 MTBE 31.5 2.4 12.7 3.2 0.8 Total 1318.8 100.0 325.5 100.0
59
Figure 2.1. Sampling sites during MCMA-2002 and MCMA-2003 field campaigns. Points
indicate the location of the sampling sites and numbers correspond to the sites listed in Table 1.
The gray scale shows the orography of the Valley of Mexico and the black contour limits the
Federal District.
60
Figure 2.2. Time series of benzene (a) and toluene (b) mixing ratios measured by DOAS, PTR-
MS and GC-FID. The resolution time for DOAS was 5 minutes and for PTR-MS ~30 sec.
PTR-MS points correspond to 1 minute averages. Samples collected by canisters and analyzed
by GC-FID correspond to 1 hour averages. Short term spikes were commonly observed with
PTR-MS indicating local sources.
61
Figure 2.3. Time series of C2-benzenes and C3-benzenes measured by the PTR-MS together
with GC-FID samples collected in parallel at La Merced (a, b), Pedregal (c, d) and Santa Ana
sites (e, f). PTR-MS concentrations represent hourly averages. GC-FID samples were collected
in hourly periods at La Merced and Pedregal sites, and in 3-hours periods at Santa Ana
Tlacotenco. The dashed lines indicate ±1 standard deviation of the PTR-MS concentrations.
62
Figure 2.4. Average diurnal pattern of the olefinic mixing ratio (as propylene) detected by the
FOS for 23 days during the MCMA-2003 study (black line) and for individual weeks. The gray
shadow represents the one standard deviation range, and gives and indication of the day-to-day
variability in each phase of the daily cycle.
63
Figure 2.5. VOC distribution by groups during the morning (a) and afternoon (b). Numbers in
the columns indicate the percent contribution of each VOC group to the total VOC
concentration, which is displayed at the bottom of each column in ppbC.
64
Figure 2.6. Correlations between alkenes comparing ambient data to vehicle chase samples.
Ambient data correspond to all canister samples and the dashed line indicates the regression line
for vehicular emissions.
65
Figure 2.7. Correlations between alkanes comparing ambient data to vehicle chase samples.
Ambient data correspond to all canister samples and the dashed line indicates the regression line
for vehicular emissions.
66
Figure 2.8. Correlations between aromatics comparing ambient data to vehicle chase samples.
Ambient data correspond to all canister samples and the dashed line indicates the regression line
for vehicular emissions.
67
Figure 2.9. Comparison of urban and vehicle exhaust hydrocarbon abundances relative to acetylene in (ppbC/ppbC). The closed circles indicate the median values for vehicle exhaust, while the open circles indicate the median values for urban data collected from urban sites (Pedregal, La Merced and CENICA) between 6 and 9:00 h. The gray shading encloses the 10th and 90th percentiles of the urban values.
olefins
Light alkanes
alkanes
aromatics
68
Figure 2.10. Olefins concentration measured by the FOS and calculated by the CIT model (a)
during 5 days and (b) average concentrations measured during the entire MCMA-2003 field
campaign at the CENICA site. The gray shading indicates the ±1 standard deviation range from
the FOS measurements.
69
70
CHAPTER 3
FLUX MEASUREMENTS OF VOLATILE ORGANIC COMPOUNDS
FROM AN URBAN LANDSCAPE*
3.1. Abstract
Direct measurements of volatile organic compound (VOC) emissions that include all sources
in urban areas are a missing requirement to evaluate emission inventories and constrain current
photochemical modelling practices. Here we demonstrate the use of micrometeorological
techniques coupled with fast-response sensors to measure urban VOC fluxes from a
neighbourhood of Mexico City, where the spatial variability of surface cover and roughness is
high. Fluxes of olefins, methanol, acetone, toluene and C2-benzenes were measured and
compared with the local gridded emissions inventory. VOC fluxes exhibited a clear diurnal
pattern with a strong relationship to vehicular traffic. Recent photochemical modelling results
suggest that VOC emissions are significantly underestimated in Mexico City, but for the olefin
class, toluene, C2-benzenes, and acetone fluxes measured in this work, the results show general
agreement with the gridded emissions inventory. While these measurements do not address the
full suite of VOC emissions, the comparison with the inventory suggests that other explanations
may be needed to explain the photochemical modelling results.
3.2. Introduction
Emission inventories provide the foundation for testing our understanding of atmospheric
chemistry at urban, regional and global scales. For VOC emissions, it has been a frequent
practice to adjust anthropogenic VOC emissions by factors between 1 and 4 to obtain matches
between photochemical model simulations and observations of ozone [Solomon et al., 2000].
This calibration of models using adjusted emissions can hide critical modelling issues that * Velasco, E., Lamb, B., Pressley, S., Allwine, E., Westberg, H., Jobson, T., Alexander, M., Prazeller, P., Molina, L. & Molina, M. Flux measurements of volatile organic compounds from an urban landscape. Geophysical Research Letters 32(20) L20802, (2005).
71
prohibit the development of a true picture of atmospheric chemistry. One way to address this
problem is to make direct measurements of VOC emissions and to use these data for an explicit
evaluation of gridded VOC emission inventories. We recently completed a set of
micrometeorological measurements of VOC fluxes as part of the Mexico City Metropolitan Area
2003 field campaign (April 7-29) in a densely populated district of Mexico City. We employed a
tall urban tower instrumented with a Fast Olefin Sensor (FOS), a Proton Transfer Reaction-Mass
Spectrometer (PTR-MS) and appropriate meteorological sensors. Fluxes of olefins were
measured with the FOS and computed by the eddy covariance technique (EC), while fluxes of
toluene, C2-benzenes, methanol and acetone were measured by the PTR-MS and calculated using
the disjunct eddy covariance method (DEC).
The applicability of micrometeorological techniques to a potentially inhomogeneous area,
such as a city with diverse emission sources (motor vehicles, industrial process, solvents
evaporation, food cooking, etc.), is confined to conditions, such that the tower height exceeds the
blending height at which the small scale heterogeneity merge into a net exchange flux above the
city. The left side of Figure 3.1 illustrates how the emissions from different sources are blended
by the turbulence produced in the roughness sub-layer by the influence of buildings, trees, and
the urban surface heating. Instruments must be mounted in the constant flux-layer, which is at
least twice the mean height of the roughness elements [Grimmond and Oke, 1999], to measure
fluxes fully representative of the underlying surface. Our measurements fulfilled both
constraints with a 37 m tower, a height 3 times the height of surrounding buildings and below a
tenth of the minimal atmospheric boundary height in Mexico City (~400 m).
3.3. Experimental methods
To instrument the flux tower, a 3-d sonic anemometer was attached to a 3 m boom near the
top of the tower. Signal/power cables and a Teflon sampling line were run from the sensors to a
shelter on the roof-top. Inlet lines to the FOS and PTR-MS were connected to the sampling line
near the roof-top instrument shelter. The lag time for sampling at the shelter was determined
72
from a covariance analysis between the vertical wind speed and the FOS signal. In operation, the
flux system collected data at 10 Hz. The data were used to calculate 30 min average fluxes.
The EC method calculates the flux of a trace gas (Fχ) as the covariance between the
instantaneous deviation of the vertical wind velocity (w’) and the instantaneous deviation of the
trace gas (cχ’) for time periods no longer than one hour [Aubinet et al., 2000]. The DEC method
is similar to EC, but with a slower sampling rate, and therefore the flux is calculated from a
smaller subset of samples.
The quality of flux measurements is difficult to assess, because there are various sources of
errors that range from failure to satisfy any of a number of theoretical assumptions to failure of
the technical setup. Systematic and random errors are estimated to restrict the accuracy of an
individual turbulent flux measurement to within 10% to 20% [Montcrieff et al., 1996]. It has
been suggested there are certain plausibility criteria that EC flux methods should fulfil [Aubinet
et al., 2000]. We investigated three criteria: the statistical characteristics of the raw
instantaneous measurements, the frequency resolution of the eddy covariance system through the
spectra and cospectra of the measured variables, and the stationarity of the measurement process.
These criteria were satisfied more than 70% of the time for our flux measurements. Details are
given in the auxiliary material.
Additional uncertainties in the olefins flux are due to analytical uncertainties. The FOS is a
chemiluminescent detector calibrated with propylene, but known to have different sensitivities to
a variety of olefins [Guenther and Hills, 1998]. We evaluated the FOS response and found that,
on average, 52% of the olefins detected by the FOS remained unknown. Additional analysis
would be needed to identify these unknown species. However, we can affirm that the FOS
measures a mix of VOC responding as propylene and can be used to provide a continuous and
fast response measurement of the olefinic VOC mix in an urban atmosphere.
The PTR-MS is a more specific sensor compared to the FOS. The PTR-MS detection is
based on mass spectroscopy of ions produced in proton-transfer reactions of H3O+ with selected
aromatic and oxygenated VOCs [Lindinger et al., 1998]. However, the fluxes of these species
73
involve uncertainties due to the use of the DEC technique. The PTR-MS operated with a
resolution varying from 1.2 to 3.6 s. Although DEC gives more time to process the samples, it
increases the statistical uncertainty of the fluxes. We evaluated this uncertainty and found the
potential error to range from 9% to 40%, with a median of 30% due to the longer time resolution.
The height of the tower in conjunction with the surface roughness and atmospheric stability
determines the footprint (upwind source area) of the measured flux. For an inhomogeneous
surface, such as an urban area, the measured signal depends on which parts of the surface have
the strongest influence on the sensor, and thus the location and size of the source footprint. To
evaluate the footprint we applied a hybrid Lagrangian model [Hsieh et al., 2000] to all of the 30-
minute periods from the campaign. For a footprint that encompasses 80% of the total flux, the
longest footprints (~ 5 km) correspond to stable conditions, which prevail at nighttime; while the
smaller footprints (500 m) correspond to unstable conditions, characteristic of morning and early
afternoon (Figure 3.1). The average footprint was 1.3 km, which approaches the grid scale (i.e. 2
km x 2 km) of the emissions inventory.
3.4. Results and discussion
The diurnal profile of olefin fluxes in Figure 3.2 shows that fluxes remained positive during
our study, which indicates that the urban surface is always a net source of olefins. The highest
fluxes occurred after sunrise, between 6:30 and 8 a.m., and lower fluxes were observed during
the night. Fluxes ranged from zero to 1.4 µg m-2 s-1, with an average of 0.36 µg m-2 s-1. The
morning peak coincided with the rush hour. Olefin fluxes were lower on the weekends than on
weekdays, and on weekends the morning rush hour peak was not observed. On average, the daily
flux on weekdays was 21% higher than during weekends. This effect is directly related to
changes in vehicular traffic and to industrial and commercial activities.
In an urban area, it is important to determine the magnitude of the fluxes as a function of the
upwind direction to identify the strongest sources. Figure 3.3 shows the olefin flux footprint
superimposed on a map of the neighborhood. Fluxes were consistently lower when winds were
74
from areas with fewer streets, such as north of the tower site. A high density of streets indicates
a residential area, while regions with few streets are occupied by factories and stores. Thus, we
can infer that residential areas and streets are the major contributor to the olefin flux in this
section of the city. The largest olefin fluxes were associated with winds from the east and
southeast directions. These sectors are densely inhabited and include the avenue #3 in Figure
3.3, the street with the highest traffic near the measurement site.
Figure 3.4 shows the fluxes of methanol, acetone, toluene and C2-benzenes. All showed
similar diurnal patterns, with highest fluxes during the morning and lowest fluxes during night.
In general, fluxes for these species were always positive. Fluxes of acetone were negative in
early morning, which suggests that deposition of acetone may occur before sunrise. The average
flux for each of these species was lower than the average olefin flux. The mean daily flux of
0.29 µg m-2 s-1 for methanol was 19% lower than the mean olefin flux. For toluene, the mean
daily flux was 0.23 µg m-2 s-1, while mean fluxes of acetone and C2-benzenes were 0.11 and 0.13
µg m-2 s-1, respectively. The spike observed with the olefin fluxes after sunrise was not recorded
for these species; this suggests that they may arise from different sources. Fluxes of these four
species increased after sunrise at 7 a.m., remained relatively constant during the morning,
decreased at noon, and then remained constant during the afternoon. On average, the fluxes for
these compounds were reduced by 53% in the afternoon relative to the morning. In some cases,
these decreases were related to wind direction variations, when the upwind changed from highly
inhabited areas before noon to areas mainly occupied by factories and stores in the afternoon.
We compared our flux measurement to the most recent emissions inventory derived for air
quality modelling [West et al., 2004]. The annual emissions inventory developed for Mexico
City was gridded using landuse, traffic pattern and industrial data, as well as population,
commercial, or other distributions, as appropriate. Speciation of the VOC emissions for the
SAPRC-99 chemical mechanism was primarily based upon Mexican emission profiles [Vega et
al., 2000]. Further details for this inventory are presented in the auxiliary material. The diurnal
patterns for calculated and measured olefin fluxes for the tower location are in good agreement
75
as shown in Figure 3.2. The measured flux of olefins was slightly higher than predicted by the
emissions inventory during early morning hours. Then, during the rest of the day, the inventory
emissions exceeded the measured fluxes by no more than 30%. The only significant discrepancy
is a smaller morning rush hour peak in the emissions inventory compared to the peak observed in
the measured fluxes. Thus, the results for this site and for the olefin class of VOC do not support
the idea that the inventory is underestimated by a factor of 3 as suggested in the modelling study
of West et al. [2004]. For toluene, we found that estimated emissions were generally greater than
the measured average flux, but the inventory emissions were within one standard deviation of the
measured value for most of the diurnal period. For C2-benzenes, we observed relatively poor
agreement between measured fluxes and the reported emissions. During the entire day,
emissions in the inventory were more than 2 times higher than the measured flux profile.
Finally, for acetone, we found that the inventoried rates were within one standard deviation of
the measured average flux throughout the diurnal period. Methanol was not specified in the
emissions inventory so a comparison was not possible.
These comparisons indicate that inventoried olefin, toluene, and C2-benzene emissions agree
with or exceed measured fluxes in this district of Mexico City. If this is true for other VOC
classes and in other sections of the city, it implies that factors other than an underestimation of
VOC emissions may be causing poor agreement between modelled and measured ozone levels.
At the same time, it is important to note that the species we measured do not account for a
majority of the VOC emissions which tend to be dominated by alkane species. For this reason, it
is important to extend these direct VOC flux measurements to include other compounds and to
also make measurements in other parts of the city.
During the study, traffic counts were performed at two intersections (roads 1 and 2; roads 2
and 3 in Figure 3.3). The typical morning and afternoon traffic peaks in urban areas were not
identified at either intersection; instead an abrupt increase in traffic was observed at 6 a.m., with
constant traffic counts during the day and, declining counts at night. Additionally, fluxes of CO2
were also measured from the tower. A comparison of the diurnal pattern of traffic counts against
76
the fluxes of olefin and CO2 showed a significant relationship. The average diurnal ratio
between fluxes of olefins and CO2 was 0.85, suggesting that both species are emitted by the same
sources. If we accept this hypothesis and assume that 61% of CO2 emissions are due to
transportation as indicated in Sheinbaum et al. [2000], we find that transportation sources
account for 46% to 76% of the olefin flux. This range coincides with the official emissions
inventory [SMA-GDF, 2004] and a recent study based upon a chemical mass balance model
[Vega et al., 2000]. For the modeling grid cell where the flux tower was located, the distribution
of sources in the emissions inventory shows that mobile sources contribute 59% of the emissions
for both olefins and C2-benzenes, and 45% and 9% of the emissions for toluene and acetone,
respectively. Emissions from painting buildings, cleaning surfaces, and solvent consumption are
the most important acetone emission sources. This distribution of source types is similar to other
sections of the city dominated by poor neighbourhoods (~70% of Mexico City).
3.5. Concluding remarks
Micrometeorological techniques have not previously been used in urban areas to measure
VOC fluxes because of the lack of fast-response VOC sensors and the uncertainties involved
with flux measurements in a city, where the spatial variability of surface cover, emission sources
and roughness is high. However, in this field study, we have demonstrated the use of EC and
DEC techniques to perform VOC flux measurements in an urban area using state of the art VOC
sensors. The capability to evaluate emissions inventories using micrometeorological techniques,
as we have described in this work, is a valuable, new tool for improving air quality management.
Results from these direct flux measurements reduce the degrees of freedom available to match
typical photochemical grid modeling results with observations and thus suggest that unresolved
uncertainties should be sought in other aspects of urban atmospheric chemistry.
77
3.6. Acknowledgements
This study was supported by the Integrated Program on Urban, Regional and Global Air
Pollution from the Massachusetts Institute of Technology and the Metropolitan Commission of
Environment of Mexico City. The authors thank the assistance of the National Center for
Environmental Research and Training as host of this study. The authors are also grateful to
Chuck Kolb of Aerodyne for comments on the manuscript, and Miguel Zavala and Benjamin
Foy of MIT, who supplied the emissions data.
3.7. References
Aubinet, M. et al. (2000), Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology. Advances in Ecological Research, 30, 113-175.
Grimmond, C. S. B., and T. R. Oke (1999), Aerodynamic properties of urban areas derived from analysis of surface form, J. Applied Meteor., 38, 1262-1292.
Guenther, A. B., and A. Hills (1998). Eddy covariance measurement of isoprene fluxes. J. Geophys. Res., 103(D11), 13145-13152.
Hsieh, C.I., G. Katul, and T. Chi (2000), An approximate analytical model for footprint estimation of scalar fluxes in thermally stratified atmospheric flows. Advances in Water Resources, 23, 765-772.
Lindinger, W., A. Hansel, and A. Jordan (1998), On-line monitoring of volatile organic compounds at pptv levels by means of proton-transfer-reaction mass spectroscopy (PTR-MS): Medical applications, food control and environmental research. Int. J. Mass Spectrom., 173, 191-241.
Montcrieff, J. B., Y. Malhi, and R. Leuning (1996), The propagation of errors in long-term measurements of land-atmosphere fluxes of carbon and water, Global Change Biology, 2, 231-240.
Sheinbaum, C., L. V. Ozawa, O. Vasquez, and G.Robles (2000), Inventario de emisiones de gases de efecto invernadero asociados a la producción y uso de energía en la Zona Metropolitana del Valle de México. Reporte Final del Grupo de Energía y Ambiente, Universidad Nacional Autónoma de México, México, D.F.
SMA-GDF (2004), Inventario de emisiones a la atmósfera. Zona Metropolitana del Valle de México, Secretaria del Medio Ambiente del Gobierno del Distrito Federal, México D.F.
78
Solomon, P. A., E. B. Cowling, G. M. Hidy, and C. S. Furiness (2000), Comparison of scientific findings from major ozone field studies in North America and Europe. Atmos. Environ., 34, 1885-1920.
Vega, E., V. Mugica, R. Carmona, and E. Valencia (2000), Hydrocarbon source apportionment in Mexico City using the chemical mass balance receptor model. Atmos. Environ. 34, 4121-4129.
West, J. J., M. A. Zavala, L. T. Molina, M. J. Molina, F. San Martini, G. J. McRae, G. Sosa, and J. L. Arriaga-Colina (2004), Modeling ozone photochemistry and evaluation of hydrocarbon emissions in the Mexico City Metropolitan Area, J. Geophys. Res., 109, 19312, doi: 10.1029/2004JD004614.
79
Figure 3.1. Left side: emissions from different sources are blended by the turbulence generated
in the roughness sub-layer, producing a net exchange flux in the constant flux layer where the
flux is measured. Right side: fraction of the flux measured (F/S0) at the tower height versus the
upwind distance or effective source footprint (x). F is the flux and S0 the source strength.
80
Figure 3.2. Average diurnal pattern of olefin fluxes predicted and measured for the 23-day
study, and for weekdays and weekends. The dashed line indicates the sum of olefinic VOC
emissions reported in the emissions inventory for the grid cell where the flux system was located.
The grey shadow represents ±1 standard deviation from the total average.
81
Figure 3.3. Olefin flux distribution as a function of the upwind direction during the entire study.
The contour indicates the footprint that is estimated to encompass 80% of the total flux. Flux
intensity is shown in grey scale.
82
Figure 3.4. Average diurnal patterns of the flux of (A) methanol, (B) acetone, (C) toluene and
(D) C2-benzenes measured by the PTR-MS during ten days of the study. The grey shadows
represent ±1 standard deviation from the total averages.
83
3.A. Auxiliary material
3.A.1. Evaluation of emission inventories
Generally, emissions inventories are evaluated through ambient measurements. The most
direct evaluation has been to compare measured concentrations of NOx and VOC with model
results [West et al., 2004; Sillman, 1999; Trainer et al., 2000]; using this approach input
emissions in models are modified to obtain the closest possible agreement with ambient
measurements, with the assumption being that all errors come from the emissions inventory and
not from model or measurements uncertainties. Other evaluation methods are based on
measured ratios between species [Goldan et al., 1995]; for example, the morning VOC/NOx ratio
has been demonstrated to be useful for this purpose because chemical losses are relatively small
during the morning and ambient concentrations reflect the ratio of directly emitted species
[Sillman, 1999; Arriaga-Colina et al., 2004]. The chemical mass balance receptor model has also
been successfully applied to evaluate and improve emissions inventories by identifying the
relative contribution of different sources to measured VOC fingerprint [Watson et al., 2001]. For
the case of emissions inventories from mobile sources, emissions have also been evaluated using
open path sensors positioned across roads and from tunnel studies. Tunnel studies have proved
to be a useful tool to measure emissions from in-use vehicles operating under real-world
conditions [Sawyer et al., 2000].
3.A.2. Micrometeorological techniques to measure fluxes of trace gases
Micrometeorological techniques have been successfully used to measure biogenic
hydrocarbon emissions from woodlands and agricultural fields. Biogenic VOC (BVOC) fluxes
have been measured by techniques such as surface layer gradient, relaxed eddy accumulation
(REA), EC and DEC [Guenther et al., 1996; Westberg et al., 2001; Karl et al., 2002]. The first
two are indirect measurements of flux that rely on empirical parameterizations. EC is a direct
flux measurement technique that requires a sampling rate of approximately 5 Hz. Fast-response
sensors have recently become available so that EC methods can be used to measure VOC fluxes
84
above a forest canopy [Guenther and Hills, 1998] and over crop lands [Shaw et al., 1998]. The
capabilities of the PTR-MS have been expanded to allow direct flux measurements of a number
of BVOC through DEC [Karl et al., 2002; Rinne et al., 2001]. This method calculates fluxes
using measurements at a lower frequency compared to EC; the sample interval can be as long as
30 s. This makes it possible to use relatively slow sensors for flux measurements. In urban
areas, micrometeorological approaches have previously been applied to measure fluxes of
momentum, heat and CO2 [Grimmond et al., 2004; Oke et al., 1999; Moriwaki and Kanda, 2004;
Soegaard and Møller-Jensen, 2003; Nemitz et al., 2002], and aerosols [Dorsey et al., 2002].
3.A.3. Field experiment
The flux measurements were conducted in a suburb at the southeast of Mexico City
(19º21’29’’ N, 99º04’24’’ W) in spring 2003. Here we describe the site, equipment used and
details of the post-processing of the data.
3.A.3.1. The CENICA super site
The CENICA super site is located in the Iztapalapa suburb, which is the suburb with the
highest population in Mexico City (1,771,673 inhabitants) and highest density (12,000
inhabitants km2) [INEGI, 2000]. Mexico City was selected for this experiment because it is a
megacity known for severe photochemical air pollution problems, and because the urban VOC
concentrations are among the highest in the world [Arriaga-Colina et al., 2004]. VOC emissions
from combustion sources are accentuated by the high elevation of the city (2240 m.s.l.), which
makes combustion processes less efficient. VOC evaporative emissions from a variety of
sources such as cleaning, painting, and industrial processes are large due to the subtropical
weather, temperatures of over 20ºC and intense solar radiation all the year round. Finally, the
main contributor to high VOC concentrations in Mexico City is the extensive presence of aged
industrial operations and a relatively old vehicle fleet.
85
A 25 m tower was installed on the rooftop of the 12 m tall CENICA laboratory. The
instruments were mounted at a height 3 times taller than the surrounding buildings, this was
assumed to be a sufficient height to be in the constant flux layer. The flux measurements were
representative of the local scale (102 – 104 m), and integrated different neighbourhoods, with
different combinations of built and vegetated cover.
3.A.3.2. The study period
The measurements of fluxes of olefins, CO2 and energy were performed during 23 days
during the dry season in April 2003 (days of the year: 97 – 119). The study period included the
Holy Week (April 14 – 20), a period in which the vehicular traffic is reduced as many of the city
residents leave for the holiday period. By taking measurements before, during and after this
period, we expected to obtain data to help determine the influence of vehicular emissions upon
atmospheric pollution. Fluxes of olefins were measured continuously with a FOS, while fluxes
of methanol, acetone, toluene and C2-benzenes were measured using a PTR-MS during selected
days, including a continuous period between April 9 and April 16. During the other portions of
the study, the PTR-MS was operated in a VOC scan mode to help identify the full range of VOC
present in the Mexico City atmosphere.
This experiment was part of a large field campaign conducted in Mexico City (MCMA-2003),
whose main objectives were to update and improve the Mexican emissions inventory, as well as
to improve the current knowledge of the chemistry, dispersion and transport processes of
pollutants emitted to the atmosphere.
3.A.3.3. Instrumentation
To instrument the flux tower (Figure 3.A.1), a 3 m boom was attached to the tower and a three
axes sonic anemometer (Applied Technologies, Inc., model SATI-3K), an open-path infrared gas
analyzer (IRGA) to monitor CO2 and water vapour, were attached at the end of the boom. The
length of the boom was enough to minimize the effects of flow distortion from the tower, and the
86
sensors were arranged to be as aerodynamic as feasible. Signal/power cables were run from the
sensors to a shelter on the roof-top where the pc data acquisition system was operated through
LabView software specifically designed for this experiment. A Teflon sampling line (5/8 inch
O.D.) was positioned from the boom down the tower to a large pump. Inlet lines to the FOS and
PTR-MS were connected to the sampling line near the roof-top instrument shelter. The lag time
for sampling at the shelter was determined from a covariance analysis between the vertical wind
speed and raw data from the FOS. The covariance was calculated for 0 to 100 s lags; the
maximum correlation occurred for a lag time of 4.3 s. In operation, the flux system collected
data at 10 Hz. The data were used to calculate 30 min average fluxes using the EC mode for the
olefins, CO2 and heat fluxes, and the DEC mode for the fluxes evaluated through the PTR-MS.
3.A.3.4. The Fast Olefin Sensor
The FOS is a Fast Isoprene Sensor calibrated with propylene instead of isoprene. Detection is
based on chemiluminescent reaction between alkenes and ozone, and the detector responds with
different sensitivity to a variety of olefins. Also there are reactions between ozone and other
trace gases that could be potential interferences. Instrument response factors for a number of
compounds have been previously reported [Guenther and Hills, 1998]. To analyze the FOS
response in the atmosphere of Mexico City we evaluated the sensitivity of 5 olefins (propylene,
ethylene, isoprene, 1-butene and 1-3 butadiene) and nitric oxide. All of the olefins exhibited a
response. Nitric oxide gave no response in this test. Sensitivities for these species are shown in
Table 3.A.1 with their corresponding relative sensitivities to propylene and FOS responses to
average concentrations measured during selected days throughout the campaign between 6 and
10 am from the sampling line of the eddy covariance system by a canister sampling system. We
found that the FOS has a more sensitive response to 1-3 butadiene and isoprene than to
propylene; however, their ambient concentrations were much lower than the ambient levels of
propylene. In comparison, species with a lower sensitivity than propylene, but with high
concentrations in urban atmospheres (e.g. ethylene) can produce large signals. Figure 3.A.2
87
compares the FOS response to the sum of olefins as measured simultaneously with the canister
sampling system. Results suggest that generally the total olefins level detected by the FOS is
larger than the sum of identified olefins from canister samples. In only 3 of the 21 sample
periods compared was the FOS response less than the sum of olefins measured in canister
samples. With these periods removed, the ratio between the sum of olefins measured by
canisters and the FOS signal shows a median of 48%. This indicates that 52% of olefins detected
by the FOS remain unknown. Additional analysis is needed to identify these unknown species.
During the campaign the FOS was calibrated 3 times per day using propylene as standard
(Scott Specialty Gases, 10.2 ppm, ±5% certified accuracy). The linear slope of instrument
response versus propylene concentration and the zero level exhibited relatively little drift during
the study period, 14% and 9%, respectively. Their possible influence in the assignment of
absolute ambient propylene concentrations was corrected through frequent calibrations.
3.A.3.5. The Proton Transfer Reaction Mass Spectrometer
The Proton Transfer Reaction Mass Spectrometry is a novel technique capable of accurate,
selective, and fast-response measurement of a subset of important VOC. This subset includes
hydrocarbons from anthropogenic origin, such as aromatic and oxygenated species [de Gouw et
al., 2003]. The PTR-MS has been described in detail elsewhere [Lindinger et al., 1998]. Briefly,
H3O+ ions, which are produced in a hollow cathode discharge, are injected into a conventional
ion drift tube. The air to be analyzed acts as the buffer gas and is continuously introduced at the
front end of the drift tube. H3O+ ions will only react with trace components in the air having
higher proton affinities than H2O:
OHRHOHR k23 +⎯→⎯+ ++ (1)
The major constituents of air, N2, O2, CO2, CH4 and noble gases do not react with H3O+. A
small portion of the flow through the reaction chamber is sampled by a quadrupole mass
spectrometer where the RH+ and H3O+ ions are mass filtered and detected by an ion multiplier.
88
The density of analyte R can be determined from the H3O+ and RH+ count rates, the rate constant
for the ion-molecule reaction, a fixed reaction time calculated from known mobility values of
H3O+ in air, and the relative ion transmission efficiencies. In practice we calibrated the
instrument sensitivity in the field using a multi-component gas standard that contained the
analytes reported here. The calibration gas was diluted with a humid zero air flow and a
multipoint calibration curve determined over the 1 to 50 ppbv range. Calibrations were
performed every 2–3 days. The instrument background was automatically recorded every 3 hours
by switching the sample flow to a humid zero air stream. Zero air was continuously generated
by passing ambient air through a Pt-catalyst trap heated to 300 ºC. Background count rates were
subtracted from the ambient data.
We operated the PTR-MS in a selective ion mode measuring up to 6 compounds of interest
with dwell times of 0.2 or 0.5 s per compound to produce a disjunct time series with a resolution
between 1.2 and 3.6 s, depending on the number of analyzed compounds. Four species, namely
methanol (m/z 33), acetone (m/z 59), toluene (m/z 93) and C2-benzenes (m/z 107) were targeted
for this study. The m/z 107 contains contributions from ethyl benzene, three xylene isomers, and
benzaldehyde [de Gouw et al., 2003]. The signal at m/z 59 is a composite of acetone and
propanal. Based upon previous measurements of acetone and propanal in urban areas, acetone is
likely to dominate (~90%) the response at this mass. The PTR-MS data was recorded on a
separate computer from the computer used for the other sensors. The meteorological raw data
and the data from the PTR-MS were continuously synchronized using a flag sent by the PTR-MS
acquisition system.
3.A.3.6. The open-path InfraRed Gas Analyzer (IRGA)
The IRGA is an instrument specially designed for fast response measurements of CO2 and
H2O fluctuations. A detailed description of it is given elsewhere [Auble and Meyers, 1992].
This instrument is quite robust, and we found that the response is relatively constant. For CO2 it
was calibrated twice per week using two CO2 standard gas mixtures (327 and 402 ppm). We
89
compared the water vapor response to the Vaisala relative humidity sensor on the tower as a
basis for calibrating the H2O signal.
3.A.4. Eddy covariance flux technique
The flux of a trace gas (Fχ) is calculated according to the EC technique as the covariance
between the instantaneous deviation of the vertical wind velocity (w’) and the instantaneous
deviation of the trace gas (cχ’) from their 30 minutes mean (equation 2). Fundamental aspects of
EC have been widely discussed elsewhere [McMillen, 1988; Aubinet et al., 2000].
∫−==
2
1
)(')('1''12
t
t
dttctwtt
cwF χχ (2)
Because an urban surface is not flat and smooth, the data from the eddy covariance system is
post-processed using archived 30 min raw data files to test that each period meets requirements
for stationarity. Post-processing includes the following steps:
1) Convert raw data signal to scientific units, apply FOS and IRGA calibration coefficients,
remove hard spikes and identify data gaps. Hard spikes can be caused by random electronic
spikes or sonic transducer blockage (e.g. during precipitation). Data gaps may occur as a result
of system breaks for routine maintenance or calibration, inadequacy of the meteorological
conditions or complete system failures. A period is rejected if it does not fulfill 83% of the
readings.
2) Identify and remove soft spikes, large short-lived departures from the period means [Schmid
et al., 2000].
90
3) Perform coordinate rotation on three-dimensional velocity components. Its aim is to eliminate
errors due to sensor tilt relative to the terrain surface or aerodynamic shadow due to the sensor or
tower structure [Aubinet et al., 2000].
4) Calculate the mean and standard deviation for each variable.
5) Remove means from each signal to obtain fluctuations (prime quantities).
6) Apply a low pass filter to all the variables processed to eliminate the presence of a possible
trend in the 30 min time series. We used a recursive digital filter [Kaimal and Finnigan, 1994].
7) Calculate 30 minute average vertical fluxes for momentum, sensible heat, latent heat, CO2 and
olefins. For olefins, the lag time needs to be applied.
9) Account for the effects of density fluctuations upon the fluxes, applying the Webb corrections
[Webb et al., 1980] to water vapor, CO2 and olefin fluxes.
10) Apply a correction to the olefin flux for high frequency loss due to damping of fluctuations
within the sampling tube and due to instrument response time. This is done using a low-pass
filtered heat flux [Massman and Lee, 2002].
11) Apply a second phase of quality control. All measured or derived variables (30 min
averages) are submitted to a plausibility test and are rejected if they fall outside statically defined
constraints for each variable (e.g. wind speed not to exceed 25 m s-1).
3.A.5. Disjunct eddy covariance technique
The DEC technique employs instantaneous measurements, but with a slower frequency
compared to the EC method. The flux is calculated as an average of a smaller subset of samples:
)(')('1''1
ii
N
itctw
NcwF ∑
=
== χχ (3)
In equation 3, the time interval between samples is longer than in equation 2, thus the sum in
3 is an ensemble average rather than a time average. This technique gives more time to process
91
the samples, but increases the statistical uncertainty of the fluxes. DEC does not increase the
statistical uncertainty of the flux value more than 8% if the time interval between samples is
shorter than the appropriate integral timescale [Lenschow et al., 1994].
In order to evaluate the statistical uncertainty of the fluxes processed by DEC, we recalculated
the olefin flux and sensible heat for the entire campaign from the 10 Hz raw data using sample
intervals of 0.6, 1.2, 2.4 and 3.6 s. The DEC fluxes were compared with the fluxes calculated by
EC and we found good agreement for sensible heat flux with a correlation coefficient of 0.99 and
slope of 1.00 for all sampling intervals; however, for olefins, a clear difference was found when
the sampling interval was higher than 1.2 s (see Figure 3.A.3). It seems that the origin, mixing
and reactivity of the anthropogenic VOC in the atmosphere affects the integral timescale in DEC.
Assuming similar behaviour between the olefin flux and the fluxes of the four species analyzed
by the PTR-MS, whose time interval between samples was not constant, the potential error due
to the DEC statistical uncertainties ranges from 9% to 40%, with a median of 30%. This error
can be reduced by making the time step between samples shorter or using a longer averaging
time.
The raw data to calculate the disjunct fluxes is post-processed following the same steps for
EC.
3.A.6. Footprint analysis
By definition the footprint is the contribution, per unit surface flux, of each unit element of
the upwind surface area to the measured vertical flux. The footprint term is also known as
effective fetch. Both Eulerian analytical models [Horst and Weil, 1994; Kormann and Meixner,
2001] and Lagrangian stochastic dispersion models [Leclerc and Thurtell, 1990; Hsieh et al.,
1997] have been used to investigate the footprint. To evaluate the effective fetch we used a
hybrid model based on similarity theory and Lagrangian stochastic simulations [Hsieh et al.,
2000]. The main advantage of this model is its ability to analytically relate atmospheric stability,
92
measurement height and surface roughness length to flux and footprint with relative simplicity.
The performance is comparable to complex Eulerian and Lagrangian models [Schmid, 2002].
We applied this model to the complete set of 30 min periods measured during the campaign to
determine the fraction of the flux measured as function of the upwind distance and stability
condition.
The effective fetch can be evaluated through the fraction of the flux measured (F/S0) at the
tower height versus the upwind distance (x). F is the measured flux and S0 the source strength.
Following the chosen footprint model, this ratio is calculated as:
⎟⎠⎞
⎜⎝⎛ −= −PP
u LDzxkS
F 12
0
1exp (4)
where k is the Von Karman constant, L the Monin-Obuhkov length, D and P are similarity
constants depending on the atmospheric stability:
D = 0.28; P = 0.59 for unstable conditions (ξ = zm / L < 0)
D = 0.97; P = 1 for near neutral and neutral conditions (| ξ | < 0.02)
D = 2.44; P = 1.33 for stable conditions (ξ > 0)
zu is a third characteristic length defined as:
⎟⎟⎠
⎞⎜⎜⎝
⎛+−⎟⎟
⎠
⎞⎜⎜⎝
⎛=
m
mmu z
zzzzz 0
0
1ln (5)
where z0 represents the surface roughness parameter (1 m), and zm the measurement height (zm =
ztower – d), considering a zero displacement plane of 8.4 m, approximately equal to 70% of the
urban canopy (d = 0.7(zcity)).
93
It is worthwhile to identify the effective fetch as function of the wind direction and hour of
the day; Figure 3.A.4 shows the average footprint during the entire study for selected F/S0 ratios
and for different intervals of time. At night during stable and neutral conditions the footprint
extends to longer distances than during unstable conditions during daytime. The effective fetch
is also distributed over larger distances during stable conditions; 80% of its accumulation occurs
in a radius between 1.5 and 4 km. Unstable conditions prevail during the morning and early
afternoon, and produce shorter footprints with a small range and a predominant direction
between south and southeast, encompassing a major roadway and intersection. During the
middle of the day, the prevailing winds shifted somewhat more to the south and the footprint
contracted so that emissions were measured from a smaller area with fewer mobile sources.
During late afternoon, unstable conditions also prevail, but with less frequency and strength, to
produce footprints longer than in previous hours and smaller than during the night.
3.A.7. Validity for the eddy covariance system
The quality of flux measurements is difficult to assess, because there are various sources of
errors that range from failure to satisfy any of a number of theoretical assumptions to failure of
the technical setup. We investigated three criteria: the statistical characteristics of the raw
instantaneous measurements, the frequency resolution of the eddy covariance system through the
spectra and cospectra of the measured variables, and the stationarity of the measurement process.
The largest number of rejected periods that did not fulfill at least one of the three criteria
occurred at night during stable conditions; approximately 65% of the nighttime sample periods
were excluded. However, during the day, the number of rejected periods decreased drastically.
Around noon, more than 90% of periods fulfilled all criteria.
3.A.7.1. Statistical characteristics of the raw instantaneous measurements
The quality control of raw instantaneous data is applied during the postprocessing of the
fluxes, where hard and soft spikes were identified and removed. Hard spikes can be caused by
94
random electronic spikes or sonic transducer blockage (e.g. during precipitation). They are
identified by absolute limits for each variable. We found between 1 and 5 hard spikes for each
period of 18,000 readings. Soft spikes are large short-lived departures from the period means.
We followed an algorithm used in previous flux studies [Schmid et al., 2000] to identify and
remove soft spikes; these occurred at a rate of approximately 37 soft spikes per period. In
addition to spike removal, plausibility tests were applied to reject averages that fell outside
statically defined constraints (e.g. wind speed not to exceed 25 m s-1).
3.A.7.2. Spectral and cospectral analyses
All eddy covariance systems attenuate the true turbulent signals at sufficiently high and low
frequencies. This loss of information results from limitations imposed by the physical size of the
instruments, their separation distances, the inherent time response, and any signal processing
associated with detrending or mean removal. In our system the FOS and PTR-MS were located
in a shelter at the base of the tower. Sample air was pumped through 45 m of Teflon tubing.
Figure 3.A.5 shows that there is a positive correlation between mixing ratios of olefins and
methanol with vertical wind speed at the measurement height. Methanol was selected as
representative specie of the measured species by the PTR-MS. Although the sample interval for
methanol is slower than for olefins, the same pattern exists for both species. Downdrafts
(periods with negative vertical wind speed) are related to relatively low mixing ratios, and while
updrafts are related to VOC rich air (e.g. in Figure 3.A.5 from 6 to 10 s).
Spectra and cospectra were calculated for sonic temperature (T), vertical wind speed (w),
water vapor, CO2 and olefins with a standard fast Fourier transform routine. For the first 4
parameters, we observed the expected frequency distributions in an unstable surface layer,
showing the characteristic -5/3 and -7/3 slopes for spectra and cospectra, respectively, in the
inertial subrange (range where the net energy coming from the energy-containing eddies is in
equilibrium with the net energy cascading to smaller where it is dissipated). Figure 3.A.6 shows
the spectra and cospectra between w and olefin concentration for 6 periods of 30 minutes at
95
different hours of the day, showing the expected -5/3 and -7/3 slopes in the inertial subrange up
to a frequency of 0.7 Hz. This frequency attenuation is due to a number of error sources
associated with the inevitable compromises with an eddy covariance system. In our experiment
this attenuation is mainly associated with the long tube used to pump the sample air down the
tower and the analyzer response. This lack of high frequency contribution to the fluxes can be
corrected using a low-pass filtering method [Massman and Lee, 2002].
The simplest flux correction factor based on the low-pass filter method is the ratio of the
measured sonic temperature flux )'( sw'T to the temperature flux calculated with the filtered
temperature )( 'w'Tsr designed to match the overall response (including sample line effects of the
FOS instrument), the corrected flux is estimated by equation (6). The resulting correction term
was between 0.95 and 1.2.
( ) ( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛==
''''''''
sr
scorrected Tw
TwcwcwF χχχ (6)
The same spectral correction factor was applied to the VOC fluxes measured by the PTR-MS.
The PTR-MS sampling interval was not constant during the experiment; masses were added and
dwell times adjusted to improve signal to noise. Sampling intervals thus varied between 0.6 and
3.6 s so spectral analyses were not conducted. However, the loss of flux at high frequencies is
assumed to be similar in both the FOS and PTR-MS.
3.A.7.3. Stationarity test
One of the basic assumptions of the eddy covariance technique is the stationarity of the
measuring process based upon the separation between the time scales of the rapid turbulent
fluctuations and the slow synoptic variations of atmospheric variables. There are several
techniques to detect periods where the turbulent variables are nonstationary [Dutaur et al., 1999].
We applied a stationarity test based on the basic equation of the eddy covariance technique to
96
determine fluxes (equation 2) [Aubinet et al., 2000]. The measured time series of 30 min
duration is divided into n/m = 6 intervals of 5 min. The covariance of a measured signal ξ and η
of the interval l with m = 3000 (3000 measuring values in 5 min for 10 Hz scanning, i.e. n =
18000 in 30 min) measuring value is:
⎥⎦
⎤⎢⎣
⎡⎟⎠
⎞⎜⎝
⎛⎟⎠
⎞⎜⎝
⎛−
−= ∑ ∑∑
= ==
m
k
m
kkl
m
kklklkl mm 1 11
11
1'' ηξηξηξ (7)
for the test, the mean covariance of the n/m single interval is used:
⎥⎦
⎤⎢⎣
⎡= ∑
=
mn
lllmn
/
1''
/1'' ηξηξ (8)
For practical use, if the difference between the covariances determined by equations (2) and
(8) is < 30%, the data is considered of high quality and if the difference is between 30% and
60%, the data have an acceptable quality [Aubinet et al., 2000]. We applied this test to the olefin
fluxes for each measured period during the entire study; Figure 3.A.7 compares the average flux
differences between 6 subperiods of 5 min and the 30 min averages versus their respective
atmospheric stability conditions through the Monin-Obukhov length. In this case, 82% of the
periods showed a difference lower than 60%, and 70% of the periods had flux differences lower
than 30% and, thus, met the stationarity criteria. In general, conditions of non-stationarity were
related to Monin-Obukhov lengths near zero which correspond to very unstable or stable
(depending on the sign) conditions. Periods that did not meet the stationarity criteria were not
used in the subsequent analyses.
3.A.8. Evaluation of random and systematic errors in the measured daily mean olefin flux
Although error propagation cannot be quantified accurately for flux measurements, a
sensitivity analysis on the effects of errors can help to determine flux uncertainties [Montcrieff et
al., 1996]. Thus, we evaluated the random and systematic errors on the mean daily flux. When
97
errors are random, errors in computed means diminish with increasing size of data set according
to N-1/2, where N is the number of points. Thus, random errors can be estimated by examining
the convergence of calculations of the net flux. In contrast, systematic errors are not affected by
increasing data set size, because they simply add in a linear fashion.
Figure 3.A.8a shows results of applying diverse values of percentage random error (pr) to
each half hour olefin flux of the entire data set. The horizontal line represents the mean daily
flux, 0.36 µg m-2 s-1. The symmetrical set of lines above and below this line represents the
overall random error as a function of the magnitude of pr and the given number of days. If we
assume a random error of 20% on each 30 minute mean flux for the 23 days studied, the total
random error on the net flux is 0.17 µg m-2 s-1 (47% of the net flux). If we also assume a
systematic error (ps) of 10%, we see from Figure 3.A.8b that the total systematic error on the net
flux is 0.035 µg m-2 s-1. Adding these two errors gives the total uncertainty of the net flux: 0.20
µg m-2 s-1 (57% of the total flux). This error analysis allows us to determine how many days of
measurements are required to obtain the real flux at a site. For example, considering a random
error of 20% and a systematic error of 10%, 100 days of measurements would resolve a net flux
of 0.47 µg m-2 s-1, 34% higher than the mean daily flux, while a year of measurements would
resolve a net flux of 0.43 µg m-2 s-1, at a level 22% greater than the base case. However, flux
measurements can be continued only for a finite time before seasonal trends become important.
The annual climate in Mexico City has three climatic seasons, dry warm season from March to
May, raining season from June to October and dry cold season from November to February.
Thus a seasonal evaluation of olefins fluxes in Mexico City using a 3 month period will yield
fluxes resolved to within 11% of the mean daily fluxes assuming an optimistic point of view
(random and systematic errors of 5%) and resolved to within 35% assuming large errors (random
errors of 20% and systematic errors of 10%).
98
3.A.9. Diurnal profile of olefin concentration
Figure 3.A.9 shows the diurnal profile of average olefinic VOC concentrations detected by the
FOS. This pattern is similar to what would be expected for typical pollutants emitted by mobile
sources, such as NO and CO [INE, 2000]. The highest olefin concentrations were measured at 7
am and ranged from 30 to 87 ppbv, with 58 ppbv as average. This morning peak is attributed to
the rapid increase of anthropogenic emissions into a shallow mixed layer followed by a rapid
dilution by an expanding mixed layer as the sun warms the surface. Low olefin concentrations
were observed during the afternoon, with an average of 6.6 ppbv, when dilution through the deep
mixed layer was large and emissions were reduced compared to morning periods. The diurnal
pattern was relatively constant during the entire study. There was a small difference of 6 ppbv in
the early morning peak between Holy Week (national holyday period, characterized by reduction
in traffic) and the other two measurement weeks.
3.A.10. Similarity between fluxes of VOC and CO2, and their relationship with vehicular
traffic
During the MCMA-2003 field campaign, traffic counts were performed at two intersections,
one between avenues labelled 1 and 2 (I1-2) in Figure 3.3 from the main text, and the other
between avenues 2 and 3 (I2-3). Figure 3.A.10 presents the diurnal olefin flux pattern along with
the vehicular count distributions, where the vehicular fleet was classified into four different
groups: passenger cars, taxi cabs, light trucks/buses, and heavy trucks/buses. Although both
intersections share avenue 2, I2-3 has a higher traffic density than I1-2, since avenue 3 encircles the
entire city, making it one of the busiest avenues, and one of the largest pollutant emission
sources. Several routes of “colectivos” (small buses with a capacity for 20 passengers) start from
I1-2, which can be observed in the higher number of light buses and trucks compared to I2-3.
The typical morning and afternoon urban traffic peaks do not occur at either intersection,
instead a single peak is observed during the day. This peak begins at 6 am, and extends until late
afternoon for I1-2, and into the night for I2-3. Assuming that the spike in the olefin flux profile
99
during the morning is due, in part, to the early morning release of olefins stored in the near
surface layer during the night, we can affirm that the olefin flux profile matches the vehicular
traffic density diurnal profile. Both remain constant during most of the daytime and then decline
at night. Figure 3.A.11a shows the ratio formed by the normalized fluxes of olefins and CO2
during the entire diurnal cycle. This ratio is close to one during the afternoon and part of the
night which suggests that both pollutants have the same emission sources. Additionally, linear
regressions between the olefin fluxes and traffic counts for both intersections confirm a degree of
correlation between the measured flux and the traffic density with correlation coefficients of 0.69
for I1-2 and 0.62 for I2-3.
Figure 3.A.11 also shows the ratios formed by the normalized fluxes of the four species
analyzed by the PTR-MS and CO2. For these four VOC, we did not find a very close
relationship to CO2 fluxes, as for olefins. Table 3.A.2 shows the average ratios of the normalized
fluxes between each VOC and CO2 for different periods of the diurnal cycle. The two
oxygenated species, methanol and acetone, and the two aromatic species, toluene and C2-
benzenes showed ratios above 0.89 during the morning; this suggests that they are emitted by
similar sources emitting also CO2. In the afternoon, their ratios were not higher than 0.61, and
during the night, not higher than 0.41 which indicates stronger contributions from emission
sources different than those emitting CO2. In general, acetone showed the lowest similarity to
CO2 fluxes with a diurnal average ratio of 0.47.
3.A.11. Comparison between the measured VOC fluxes and VOC emissions used for air
quality models in Mexico City
In order to compare the measured VOC fluxes with the emissions reported in the emissions
inventory, we used the corresponding VOC emissions for the grid cell where our flux system was
located, from the input data used for the most recent gas-phase modelling for Mexico City [West
et al., 2004]. These emissions were based on the annual emissions reported in the official
emissions inventory for 1998 [CAM, 2001]. This inventory was created by local government
100
authorities using bottom-up methods and emissions factors which were either measured locally
or taken from elsewhere; for mobile sources, the MOBILE5 emissions model was adapted to
account for local vehicle characteristics. The annual emissions were distributed spatially and
temporally (hourly) for each of 21 species categories in the SAPRC-99 chemical mechanism
according to data from a variety of sources. For mobile source emissions, it was used
information compiled by the Metropolitan Environmental Commission of Mexico City (CAM),
who estimated the spatial and temporal distributions separately for each pollutant on the basis of
traffic count data. For emissions from vegetation, it was used the spatial distribution modeled by
Velasco [2003], and temporal distributions estimated by Guenther et al. [2000]. Emissions from
area sources were distributed spatially for each category separately, using population,
commercial, or other distributions, as appropriate. For emissions from point sources, it was used
a database compiled by the CAM including 6230 industries. The database contains the location,
emissions, work hours, and other data that were used for estimating spatial and temporal
distributions. The VOC speciation was based on the SAPRC-99 chemical mechanism for VOC
reactivity assessment [Carter, 2000] and a standard mixture of hydrocarbons in urban
atmospheres in the United States [Jeffries et al., 1989]. The speciation was determined for each
source category using emissions profiles measured in Mexico City [Mugica et al., 1998, 2002;
Vega et al., 2000]. These profiles were adjusted to include species and source categories not
measured in Mexico City using emissions profiles from the SPECIATE database [Environmental
Protection Agency (EPA), 1993].
Emissions of species belonging to the groups OLE1 and OLE2 in the SAPRC-99 mechanism
were weighted by their apportionment to the standard mixture of hydrocarbons to these two
groups and added to the emissions of isoprene and ethene to compare the olefin fluxes.
Emissions from the emissions inventory were weighted also by their sensitive response to the
FOS (see Table 3.A.1), for C4 and C5 olefins (C4 terminal alkenes, 3-methyl-1-butene,
isobutene, 2-methyl-1-butene, trans-2-butene, cis-2-butene, and 2-methyl-2-butene) we used the
same response factor for 1-butene. For species whose sensitive factors were unknown we
101
assumed null factors (1-pentene, 1-hexene, C5, C6, C7, C8, C9, C10 and C11 alkenes,
cyclohexene, C4, C5, C6, C7, C8, C9, C10 and C11 internal alkenes, and cyclic or di-olefins).
For toluene, its apportionment to the standard mixture of hydrocarbons was weighted with the
total apportionment of species lumped in the group ARO1 of the SAPRC-99 mechanism. For
C2-benzenes, apportionments of o,m,p-xylenes were weighted to the total emissions of
hydrocarbons lumped in the group ARO2 and added to the reported emissions of ethyl benzene.
Acetone emissions were directly compared to the measured fluxes of acetone, since the SAPRC-
99 model considers a specific group for it. For methanol, no comparison was possible because
methanol emissions were not considered in the model.
Figure 3.A.11 shows the diurnal profiles of the measured fluxes and emissions of toluene, C2-
benzene and acetone. Emissions from a landfill reported for the grid where the flux tower was
located were excluded from the emissions profile of the emissions inventory, since the landfill
was outside of the analyzed footprint.
For the studied grid, the major contributor to the olefins and C2-benzenes budget according to
the emissions inventory are mobile sources, mainly gasoline vehicles, with a contribution close
to 60%. Other important sources for both hydrocarbons are painting buildings and cleaning
surfaces. For olefin emissions, leakages of LPG account for only 5%. For C2-benzenes another
important source is solvent consumption, which represents 8% of total emissions. For emissions
of toluene and acetone, the sum of emissions from area sources is higher than emissions from
mobile sources. However, emissions from gasoline vehicles account for 39% of all toluene
emissions, while for acetone mobile sources account for only 8%. Again, painting buildings and
cleaning surface represent the major area sources for these two compounds.
102
3.A.12. References
Arriaga-Colina, J.L., J. J. West, G. Sosa, S. S. Escalona, R. M. Orduñez, and A. D. M. Cervantes (2004), Measurements of VOCs in Mexico City (1992-2001) and evaluation of VOCs and CO in the emissions inventory, Atmos. Environ., 38, 2523-2533.
Aubinet, M., et al. (2000), Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology, Adv. Ecol. Res., 30, 113-175.
Auble, D.L., and T. P. Meyers (1992), An open path, fast response infrared absorption gas analyzer for H2O and CO2, Bound.-Layer Meteor., 59, 243-256.
CAM (2001), Inventario de emisiones a la atmósfera, Zona Metropolitana del Valle de México, 1998, Comisión Ambiental Metropolitana, México, D.F.
Carter, W. P. L. (2000), Documentation of the SAPRC-99 Chemical Mechanism for VOC reactivity assessment, final report to California Air Resources Board, contracts 92-329 and 95-308, Calif. Air Res. Board, Sacramento, Calif. (available at www.cert.ucr.edu/~carter)
de Gouw, J. A., P. D. Goldan, C. Warneke, W. C. Kuster, J. M. Roberts, M. Marchewka, S. B. Bertman, A. A. P. Pszenny, and W. C. Keene (2003), Validation of proton transfer reaction-mass spectrometry (PTR-MS) measurements of gas-phase organic compounds in the atmosphere during the New England Air Quality Study (NEAQS) in 2002, J. Geophys. Res., 108, 4682-4199.
Dorsey, J. R., E. Nemitz, M. W. Gallagher, D. Fowler, P. I. Williams, K. N. Bower, and K. M. Beswick (2002), Direct measurements and parameterisation of aerosol flux, concentration and emission velocity above a city, Atmos. Environ., 36, 791-800.
Dutaur, L., S. Cieslik, A. Carrara, and A. Lopez (1999), The detection of nonstationarity in the determination of deposition fluxes, in Proceedings of EUROTRAC Symposium ’98 Vol. 2, edited by P. M. Borrell and P. Borrell, pp. 171-176, WIT Press, Southampton.
Environmental Protection Agency (1993), SPECIATE, VOC/PM speciation data system, version 1.50, Environ. Prot. Agency and Radian Corp., Research Triangle Park, N. C.
Goldan, P.D., M. Trainer, W. C. Kuster, D. D. Parrish, J. Carpenter, J. M. Roberts, J. E. Yee, and F. C. Fehsenfeld (1995), Measurements of hydrocarbons, oxygenated hydrocarbons, carbon monoxide, and nitrogen oxides in an urban basin in Colorado: implications for emission inventories, J. Geophys. Res., 100, 22771-22783.
Grimmond, C. S. B., J. A. Salmond, T. R. Oke, B. Offerle, and A. Lemonsu (2004), Flux and turbulence measurements at a densely built-up site in Marseille: heat, mass (water and carbon dioxide), and momentum, J. Geophys. Res., 109, 24101-24120.
Guenther, A., et al. (1996), Isoprene fluxes measured by enclosure, relaxed eddy accumulation, surface layer gradient, mixed layer gradient, and mixed layer mass balance techniques, J. Geophys. Res., 101, 18555-18567.
103
Guenther, A., and A. Hills (1998), Eddy covariance measurement of isoprene fluxes, J. Geophys. Res., 103, 13145-13152.
Guenther, A., C. Geron, T. Pierce, B. Lamb, P. Harley, and R. Fall (2000), Natural emissions of non-methane volatile organic compounds, carbon monoxide, and oxides of nitrogen from North America, Atmos. Environ., 34, 2205– 2230.
Horst, T. W., and J. C. Weil (1994), How far is far enough? The fetch requirements for micrometeorological measurement of surface fluxes, J. of Atmos. and Oceanic Technol., 11, 1018-1025.
Hsieh, C. I., G. G. Katul, J. Schieldge, J. T. Sigmon, and K. K. Knoerr (1997), The Lagrangian stochastic model for fetch and latent heat flux estimation above uniform and nonuniform terrain, Water Resour. Res., 33, 427-438.
Hsieh, C. I., G. Katul, and T. Chi (2000), An approximate analytical model for footprint estimation of scalar fluxes in thermally stratified atmospheric flows, Adv. Water Resour., 23, 765-772.
INE (2000), Almanaque de datos y tendencias de la calidad del aire en ciudades mexicanas, INE-SEMARNAT, México, D.F.
INEGI (2000), Estados Unidos Mexicanos Resultados Prelimanares XII Censo General de Población y Vivienda, INEGI, México, D.F.
Jeffries, H. E., K. G. Sexton, J. R. Arnold, and T. L. Kale (1989), Validation testing of new mechanisms with outdoor chamber data, Volume 2: analysis of VOC data for the CBA and CAL photochemical mechanisms, final report EPA-600/3-89-010b, EPA, Research Triangle Park, NC.
Kaimal, J. C., and J. J. Finnigan (1994), Atmospheric boundary layer flows: Their structure and measurement, 289 pp., Oxford Univ. Press, New York.
Karl, T. G., C. Spirig, J. Rinne, C. Stroud, P. Prevost, J. Greenberg, R. Fall, and A. Guenther (2002), Virtual disjunct eddy covariance measurements of organic compound fluxes from a subalpine forest using proton transfer reaction mass spectroscopy, Atmos. Chem. Phys., 2, 279-291.
Kormann, R., and F. X. Meixner (2001), An analytical footprint model for neutral stratification, Bound.-Layer Meteor., 99, 207-224.
Leclerc, M. Y., and G. W. Thurtell (1990), Footprint prediction of scalar fluxes using a Markovian analysis, Bound.-Layer Meteor., 52, 247-258.
Lenschow, D. H., J. Mann, and L. Kristensen (1994), How long is long enough when measuring fluxes and other turbulence statistics?, J. Atmos. Oceanic Technol., 11, 661-673.
Lindinger, W., A. Hansel, and A. Jordan (1998), On-line monitoring of volatile organic compounds at pptv levels by means of proton-transfer-reaction mass spectroscopy (PTR-MS):
104
Medical applications, food control and environmental research, Int. J. Mass Spectrom., 173, 191-241.
Massman, W. J. and X. Lee (2002), Eddy covariance flux corrections and uncertainties in long-term studies of carbon and energy exchanges, Agric. For. Meteor., 113, 121-144.
McMillen, R (1998), An eddy correlation technique with extended applicability to non-simple terrain, Bound.-Layer Meteor., 43, 231-245.
Montcrieff, J. B., Y. Malhi, and R. Leuning (1996), The propagation of errors in long-term measurements of land-atmosphere fluxes of carbon and water, Global Change Biol., 2, 231-240.
Moriwaki, R., and M. Kanda (2004), Seasonal and diurnal fluxes of radiation, heat, water vapor, and carbon dioxide over a suburban area, J. Appl. Meteor., 43, 1700-1710.
Mugica, V., E. Vega, J. L. Arriaga, and M. E. Ruiz (1998), Determination of motor vehicle profiles for non-methane organic compounds in the Mexico City metropolitan area, J. Air Waste Manage. Assoc., 48, 1060–1068.
Mugica, V., J. Watson, E. Vega, E. Reyes, M. E. Ruiz, and J. Chow (2002), Receptor model source apportionment of nonmethane hydrocarbons in Mexico City, Sci. World J., 2, 844– 860.
Nemitz, E., K. J. Hargreaves, A. G. Mcdonald, J. R. Dorsey, and D. Fowler (2002), Micrometeorological measurements of the urban heat budget and CO2 emissions on a city scale, Environ. Sci. Technol., 36(14), 3139-3146.
Oke, T. R., R. A. Spronken-Smith, E. Jáuregui, and C. S. B. Grimmond (1999), The energy balance of central Mexico City during the dry season, Atmos. Environ., 33, 3919-3930.
Rinne, H. J. I., A. Guenther, C. Warneke, J. A. de Gouw, and S. L. Luxembourg (2001), Disjunct eddy covariance technique for trace gas flux measurements, Geophys. Res. Lett., 28, 3139-3142.
Sawyer, R. F., R. A. Harley, S. H. Cadle, J. M. Norbeck, R. Slott, and H. A. Bravo (2000), Mobile sources critical review : 1998 NARSTO assessment, Atmos. Environ., 34, 2161-2181.
Schmid, H. P., C. S. B. Grimmond, F. Cropley, B. Offerle, and H. B. Su (2000), Measurements of CO2 and energy fluxes over a mixed hardwood forest in the mid-western United States, Agric. For. Meteor., 103, 357-374.
Schmid, H. P. (2002), Footprint modeling for vegetation atmosphere exchange studies: a review and perspective, Agric. For. Meteor., 113, 159-183.
Shaw, W. J., C. W. Spicer, and D. V. Kenny (1998), Eddy correlation fluxes of trace gases using a tandem mass spectrometer, Atmos. Environ., 32, 2887-2898.
Sillman, S. (1999), The relation between ozone, NOx, and hydrocarbons in urban and polluted rural environments, Atmos. Environ., 33, 1821-1845.
105
Soegaard, H., and L. Møller-Jensen (2003), Towards a spatial CO2 budget of a metropolitan region based on textural image classification and flux measurements, Remote Sens. Environ., 87, 283-294.
Trainer, M., D. D. Parrish, P. D. Goldan, J. Roberts, and F. C. Fehsenfeld (2000), Review of observation-based analysis of the regional factors influencing ozone concentrations, Atmos. Environ., 34, 2045–2061.
Vega, E., V. Mugica, R. Carmona, and E. Valencia (2000), Hydrocarbon source apportionment in Mexico City using the chemical mass balance receptor model, Atmos. Environ., 34, 4121–4129.
Velasco, E. (2003), Estimates for biogenic non-methane hydrocarbons and nitric oxide emissions in the Valley of Mexico, Atmos. Environ., 37, 625-637.
Watson, J. G., J. C. Chow, and E. Fujita (2001), Review of volatile organic compound source apportionment by chemical mass balance, Atmos. Environ., 35, 1567-1584.
Webb, E. K., G. I. Pearman, and R. Leuning (1980), Correction of flux measurements for density effects due to heat and water vapour transfer, Q. J. R. Meteorol. Soc. 106(447), 85-100.
West, J. J., M. A. Zavala, L. T. Molina, M. J. Molina, F. San Martini, G. J. McRae, G. Sosa, and J. L. Arriaga-Colina (2004), Modeling ozone photochemistry and evaluation of hydrocarbon emissions in the Mexico City Metropolitan Area, J. Geophys. Res., 109, 19312, doi: 10.1029/2004JD004614.
Westberg, H., B. Lamb, R. Hafer, A. Hills, P. Shepson, and C. Vogel (2001), Measurement of isoprene fluxes at the PROPHET site, J. Geophys. Res., 106, 24347-24358.
106
Table 3.A.1. Sensitivities, average concentrations measured during selected days throughout the
campaign between 6 and 10 am from the sampling line of the eddy covariance system by a
canister sampling system, FOS responses to those average concentrations, and relative
sensitivities to propylene for the 6 evaluated species
Compound Sensitivity (photons ppb-1 s-1)
Average conc., 6–10 am (ppb)
Average FOS response (photons s-1) c
Relative sensitivity to propylene d
Propylene 25.4 7.50 191 1.00
Isoprene 74.7 0.304 23 2.94
Ethylene 17.7 20.7 366 0.70
1-3 butadiene 49.8 0.791 39 1.96
1-Butene 7.9 3.90 a 31 0.31
NO ~0.0 61.2 b 0 ~0.0
a As i-butene b From continuous monitoring during the entire campaign. c Average response = (sensitivity)(average conc.) + (zero value), in this table is considered a zero value equals to zero. d Relative sensitivity = (compound sensitivity) / (propylene sensitivity)
107
Table 3.A.2. Average ratios of the normalized fluxes of each analyzed VOC and CO2 for different periods of the diurnal cycle. A high ratio indicates that the VOC specie and CO2 are emitted by the same sources, while a low ratio indicates that the VOC and CO2 species are emitted by different sources
Olefins Methanol Acetone Toluene C2-benzenes
Morning (6:30-12:00)
0.70 0.89 0.96 0.92 0.95
Afternoon (12:30-19:00)
0.84 0.57 0.55 0.61 0.59
Night (19:30-6:00)
0.82 0.36 0.18 0.41 0.38
Day (6:30-19:00)
0.87 0.72 0.74 0.76 0.76
Diurnal average 0.85 0.55 0.47 0.59 0.59
108
Figure 3.A.1. Schematic diagram of the instrumented flux tower. The azimuth orientation of the
sonic anemometer is 16º from north. Dimensions are in meters.
109
Figure 3.A.2. Comparison of identified olefins weighted by their corresponding FOS
sensitivities from GC-FID measurements versus FOS measurements for 21 sample periods
during selected days throughout the campaign between 6 and 10 am.
110
Figure 3.A.3. Comparison between olefin fluxes (µg m-2 s-1) calculated using the techniques of
eddy covariance and disjunct eddy covariance with different sampling intervals (a) 0.6 s, (b) 1.2
s, (c) 2.4 s and (d) 3.6 s.
111
Figure 3.A.4. Different fractions of the measured flux (F/S0) during the entire campaign as
function of the wind direction for different intervals of time, (a) from 0:00 to 6:00 h, (b) from
6:00 to 9:00 h, (c) from 9:00 to 15:00 h and (d) from 14:00 to 21:00 h.
112
Figure 3.A.5. A 20-s trace of methanol and olefins mixing ratios plotted together with vertical
wind speed beginning at 9:40 h local time on April 11, 2003.
113
Figure 3.A.6. Spectra for olefins mixing ratio (a) and cospectra for olefins mixing ratio and
vertical wind speed (b) for 6 different periods of 30 minutes before the low-pass filter correction.
The -5/3 and -7/3 slopes indicate the theoretical slopes of the inertial subrange.
114
Figure 3.A.7. Stationarity test for olefin flux: a measured period is stationary if the average flux
from 6 continuous subperiods of 5 min is within 60% of the flux obtained from a 30 min
average. In our study 82% of the periods fulfilled this criterion.
115
Figure 3.A.8. Effects of random and systematic errors for the mean daily olefin flux. (a) Effects
due to random errors for various percentages of error. (b) Effects due to systematic errors as a
function of the error percentage. In both figures the horizontal line represents the mean daily
flux, 0.355 µg m-2 s-1.
116
Figure 3.A.9. Average diurnal pattern of olefinic concentration detected by the FOS for the
entire study (black line) and for individual weeks. The gray shadow represents ±1 standard
deviation from the total average, and gives an indication of the day-to-day variability in each
phase of the daily cycle.
117
Figure 3.A.10. Diurnal profile of olefin fluxes and traffic counts for two intersections within the
tower footprint (intersection between Av. Ermita Iztapalapa and Av. Rojo Gomez I1-2, and the
intersection between Av. Ermita Iztapalapa and Av. Anillo Periferico I2-3).
118
Figure 3.A.11. Ratios between the normalized fluxes of VOC and CO2. (a) for olefins, (b)
methanol, (c) acetone, (d) toluene, and (e) C2-benzenes.
119
Figure 3.A.12. Comparison of the diurnal profiles of fluxes of toluene (a), C2-benzenes (b), and acetone (c) with the diurnal profiles of emissions used in the CIT three-dimensional photochemical model for Mexico City. The dashed lines indicate ±1 standard deviation of the measured fluxes.
120
121
CHAPTER 4
MEASUREMENTS OF CO2 FLUXES FROM THE MEXICO CITY
URBAN LANDSCAPE*
4.1. Abstract
In a densely populated section of Mexico City, an eddy covariance flux system was deployed
on a tall urban tower to obtain direct measurements of CO2 emissions from an urban
neighborhood located in a subtropical megacity. The measured fluxes and boundary layer
conditions satisfy eddy covariance assumptions of stationarity, and co-spectral analyses of the
turbulence measurements exhibit the required boundary layer patterns for acceptable flux
measurements. Results from a field experiment conducted during April, 2003 show that the
urban surface is a net source of CO2. The CO2 flux measurements showed a clear diurnal
pattern, with the highest emissions during the morning (up to 1.60 mg m-2 s-1), and lowest
emissions during nighttime. The measured fluxes were closely correlated to traffic patterns in
the area. The mean daily flux was 0.41 mg m-2 s-1, which is similar to that observed in European
and U.S. cities.
4.2. Introduction
Urban landscapes are suspected to be major sources of anthropogenic CO2, however, there
are few direct measurements of CO2 emissions in urban areas. Micrometeorological techniques
have only recently been applied to urban landscapes, while they have been widely used to
measure fluxes of CO2, water vapor, volatile organic compounds (VOCs) and other trace gases
above vegetation (Baldocchi, et al., 2001; Schmid et al., 2000; Westberg, et al., 2001). Recent
urban studies have focused on cities from developed countries in mid-latitudes, such as
* Velasco E, Pressley S, Allwine E, Westberg H, Lamb B, 2005. Measurements of CO2 fluxes from the Mexico City urban landscape. Atmospheric Environment 39(38), 7433-7446.
122
Edinburgh (Nemitz et al., 2002), Chicago (Grimmond et al., 2002), Copenhagen (Soegaard and
Møller-Jensen, 2003), Basel (Vogt et al., 2003), Tokyo (Moriwaki and Kanda, 2004),
Vancouver (Walsh et al., 2004), and Marseille (Grimmond et al., 2004). Despite the fact that
three-quarters of the world’s 3 billion urban residents live in developing countries, there have
been few, if any, flux measurements in urban areas in developing countries. Mexico City is a
good example of an urban area in a less developed country. It is the second largest city in the
world, and it is characterized by rapid population growth with a wide range of both direct and
indirect CO2 sources, such as mobile emissions and land use changes due to urbanization.
CO2 levels in urban environments have typically been quantified through emission
inventories from estimates of fossil fuel consumption, evaluations of the amount of carbon
sequestered in urban vegetation, and short-term studies of ambient CO2 concentrations (see
Grimmond et al., 2002). For Mexico City, results of only two studies have been reported; the
first was a study during 1981 and 1982, in which CO2 was sampled at different locations in the
city and analyzed by gas chromatography (Baez et al., 1988). In the second study, CO2
concentrations were measured by a Fourier Transform Infrared (FTIR) spectrometer at the
southwest of the city during the fall of 2001 (Grutter, 2003).
As part of a large air quality field campaign conducted in Mexico City (MCMA-2003), we
deployed an eddy covariance (EC) flux system on a tall urban tower within a densely populated
section of the city to obtain direct measurements of CO2 emissions in a typical Mexican
neighborhood. In this paper, we demonstrate the applicability of the EC technique to a
heterogeneous urban landscape where the spatial variability of surface cover and roughness is
high. The results are analyzed in terms of the magnitude of the CO2 fluxes in relation to the
source footprint as a function of wind direction and in relation to vehicular activity.
123
4.3. Methods
4.3.1. Measurement site and study period
Mexico City is characterized by an enormous social difference, where approximately 10% of
the population lives in affluent neighborhoods, 20% in medium class neighborhoods and 70% in
poor neighborhoods. Overall, poor neighborhoods are formed by residential areas built with
scarce planning around original buildings, industries or services located throughout the city. The
flux measurements were performed in Iztapalapa, a poor neighborhood located in southeast
Mexico City (19º 21’ 29’’ N, 99º 04’ 24’’ W). Iztapalapa is the suburb with the highest
population in Mexico City (1.8 million inhabitants) and the highest density (12,000 inhabitants
km-2) (INEGI, 2000). Mexico City is located at 2240 m above sea level in a subtropical region
surrounded by mountains; the city experiences mild weather, temperatures of over 20ºC, and
intense solar radiation all year.
CO2 fluxes were measured for 23 days during the warm dry season in April 2003 (April 7 -
29). April is the second warmest month of the year with an average monthly precipitation rate
less than 25 mm of rain. The study period included the Holy Week (April 14 – 20), when traffic
is significantly reduced because many city residents leave for the holiday period. By taking
measurements before, during and after this period, we expected to obtain a better understanding
of the influence of the vehicular emissions upon CO2 fluxes.
The EC flux system was deployed at the CENICA super site. Surrounding topography
around this site is flat. At 2.4 km in the southeast direction, there is a small hill with a height of
160 m. This hill embraces the national park “Cerro de la Estrella” (see Figure 8). Including the
vegetation of this park, the total vegetation cover in Iztapalapa represents 16% of the total area;
however only 27% of that vegetation cover corresponds to woodland, the remaining 73%
corresponds to grasslands and shrubs (SMA-GDF, 2004). This means that the biomass of the
region is scarce and the potential for CO2 uptake from vegetation is small. In contrast, the
number of CO2 sources is large, and composed of a mix of commercial, industrial, residential
and mobile sources. The predominant land use in the 30o-280º sector is residential and
124
commercial, with buildings of 3 and 4 stories of height, and roadways of 1 and 2 lanes with a
density of ~16 km km-2. In general during the daytime, this sector is the predominate upwind
direction. At night, winds from all directions were observed, but winds blowing from the north
and northwest sectors were predominant. This sector is mainly occupied by light industries,
storage buildings and the Autonomous Metropolitan University. In this section, the number of
roadways is lower with an average density of 7 km km-2. Overall, the average building height
(zcity) in all sectors is 12 m. Most buildings are built of concrete with flat roofs of concrete or
sheet metal.
4.3.2. Instrumentation
The EC flux system was mounted on a 25 m tower on the rooftop of a building, with a total
height of 37 m, more than 3 times the height of the surrounding buildings, and of sufficient
height to be in the constant flux layer. To instrument the flux tower (Figure 1), a three-
dimensional (3-D) sonic anemometer (Applied Technologies, Inc., model SATI-3K), and a
NOAA open-path infrared gas analyzer (IRGA) were mounted at the top of the tower at the end
of a 3 m boom. The length of the boom was enough to minimize the effects of flow distortion
from the tower, and the sensors were arranged to be as aerodynamic as feasible. Signal/power
cables were run from the sensors to a shelter on the roof where the pc data acquisition system
was operated through LabVIEW (National Instruments) software specifically designed for this
experiment. In operation, the flux system collected data at 10 Hz, and the turbulence data were
used to calculate 30 min average fluxes.
The NOAA IRGA is an instrument designed for fast response measurements of H2O and CO2
fluctuations and developed by the Atmospheric Turbulence and Diffusion Division (ATDD) of
the National Oceanic and Atmospheric Administration (NOAA). Noise levels for H2O and CO2
are less than 0.1 g m-3 and 0.3 mg m-3, respectively. These levels are much lower than the
observed variations in the atmosphere of Mexico City; the standard deviation of H2O and CO2
mixing ratios on a 30 min basis were more than 2.4 g m-3 and 70 mg m-3, respectively. Auble
125
and Meyers (1992) have given a detailed description of this instrument. Although open path
instruments do not require frequent calibrations, we performed CO2 calibrations twice per week
since in a polluted urban atmosphere the mirrors and windows become dirty quickly. The CO2
calibration used two standard gas mixtures (Scott-Marrin Inc. 327 and 402 ppm ± 1%, National
Institute of Standards and Technology (NIST)). We compared the water vapor response to a
humidity sensor (Vaisala, model HMP45A) on the tower as a basis for calibration of the
instrument for water vapor.
4.3.3. Postprocessing for eddy covariance flux calculations
The flux of a trace gas (Fχ) is calculated according to the EC technique as the covariance
between the instantaneous deviation of the vertical wind velocity (w’) and the instantaneous
deviation of the trace gas concentration (cχ’) from their 30 minute means. The equation is given
below, where the over bar denotes a time-averaged quantity.
)(')('1''1
ii
N
itctw
NcwF ∑
=
== χχ (1)
Fundamental aspects of EC have been widely discussed elsewhere (e.g. McMillen, 1988;
Aubinet et al., 2000). Briefly the post-processing includes the following steps:
1) Convert raw data signals to scientific units, apply IRGA calibration coefficients, remove hard
spikes and identify data gaps. Hard spikes can be caused by random electronic spikes or
sonic transducer blockage (e.g. during precipitation). Data gaps may occur as a result of
system breaks for routine maintenance or calibration, inadequacy of the meteorological
conditions or complete system failures. We found between 1 and 5 hard spikes for each
period of 18,000 readings. A period is rejected if it does not successfully fulfill 83% of the
readings.
126
2) Identify and remove soft spikes which are large short-lived departures from the period means.
We followed the algorithm proposed by Schmid et al. (2000). These spikes occurred at a rate
of approximately 37 per period.
3) Perform coordinate rotation on three-dimensional velocity components. The aim is to
eliminate errors due to sensor tilt relative to the terrain surface or aerodynamic shadow due to
the sensor or tower structure (Aubinet et al., 2000).
4) Calculate the mean and standard deviation for each variable.
5) Remove means from each signal to obtain fluctuations (prime quantities).
6) Apply a low pass filter to all the processed variables to eliminate the presence of a possible
trend in the 30 min time series. We used a recursive digital filter (Kaimal and Finnigan,
1994).
7) Calculate 30 minute average vertical fluxes for momentum, sensible heat, latent heat, and
CO2.
8) Account for the effects of density fluctuations upon the fluxes, applying the Webb corrections
(Webb et al., 1980) to water vapor and CO2.
9) Apply a second phase of quality control. All measured or derived variables (30 min averages)
are submitted to a plausibility test and are rejected if they fall outside statically defined
constraints for each variable (e.g. wind speed not to exceed 25 m s-1).
The quality of flux measurements is difficult to assess, because there are various sources of
errors, ranging from failure to satisfy a number of theoretical assumptions to failure of the
technical setup. Conflicts with the assumptions of the eddy covariance technique to estimate
fluxes arise under certain meteorological conditions and site properties. As these effects cannot
be quantified solely from eddy covariance data, a classical error analysis and error propagation
will remain incomplete. Instead, Aubinet et al. (2000) suggest an empirical approach to
determine whether the fluxes meet certain plausibility criteria. Besides the statistical
characteristics of the raw instantaneous measurements, we investigated the frequency resolution
127
of the eddy covariance system through the spectra and cospectra of the measured variables, and
through the stationarity of the eddy flux process.
4.3.4. Spectral and cospectral analysis
An EC system attenuates the true turbulent signal at sufficiently high and low frequencies
due to limitations imposed by the physical size of the instruments, their separation distances,
their inherent time response, and any signal processing associated with detrending or mean
removal (Massman and Lee, 2002). Inspection of the spectra and cospectra of the measured
variables helps to determine the influence of these attenuations. Power density spectra of CO2
concentration and ambient temperature (T) are shown in Figure 4.2.a, while cospectra of vertical
velocity with CO2 concentration and T are shown in Figure 4.2.b. These spectra and cospectra
were computed via the Fast Fourier Transform (FFT) method. Each spectrum and cospectrum
was normalized by the corresponding average scalar or covariance to facilitate the comparison
between CO2 and T, and they were plotted against the normalized frequency, f (zm / u), where zm
is the measurement height and u is the mean wind velocity. The height zm is calculated as the
difference between the tower height and the zero displacement plane, which is based on rule-of-
thumb estimates (where zd = 0.7zcity, Grimmond and Oke, 1999) as 8.4 m. We use logarithmic
scales in both axes to emphasize power-law relationships.
Both spectra and cospectra of CO2 and T show similar patterns at high frequencies, f (zm / u)
> 2. At that same frequency range, the spectra show the expected -5/3 slope in the inertial
subrange, which is the range where the net energy coming from the energy-containing eddies is
in equilibrium with the net energy cascading to smaller scales where it is dissipated. The
cospectrum provides an additional mean to identify frequency-dependent behavior in an EC
system that can lead to systematic errors in flux measurements. This is because the cospectrum
reflects the phase relationship between two variables and their respective amplitudes as a
function of frequency. CO2 concentration and T were measured by independent instruments;
128
therefore, the correspondence between their cospectra with vertical wind velocity indicates that
there is no systematic phase shift and no cospectral distortion.
The spectra and cospectra results clearly show that our flux system is fully capable of
measuring turbulence fluxes of CO2 via the EC mode in an urban environment. However,
sampling periods that did not fulfill the stationarity requirements (described below) generally did
not meet the spectral and cospectral analysis criteria. Those sampling periods were not omitted
from further analysis based on the spectral and cospectral analysis, but they were discarded
based on the stationarity test which is easier to apply in an objective manner.
4.3.5. Stationarity test
The applicability of an urban flux tower is confined to stationarity conditions, such that the
measurement height exceeds the blending height at which the small scale heterogeneity merge
into a net exchange flux above the city. One criterion for stationarity is to see if the difference
between the flux obtained from a 30 min average and the average of fluxes from 6 continuous
subperiods of 5 min from that same period of 30 min is less than 60% (Aubinet et al., 2000).
When the turbulent variables are not stationary, the fluxes could be suspicious, and therefore
they should not be considered for subsequent analyses. In our study, the stationarity condition
was fulfilled in 74% of the half hour periods. Figure 3 compares the average flux differences
between 6 subperiods of 5 min and the 30 min averages versus their respective atmospheric
stability conditions through the stability index (ξ) during the entire campaign. If ξ < 0
atmospheric conditions are unstable and if ξ ≥ 0 conditions are stable where ξ = zm / L and L is
the Monin-Obukhov length.
The stationarity test can also eliminate sampling periods affected by short term horizontal
advection from local sources. Short term advection is produced by plumes emitted during short
periods of time, or plumes transported during changing wind conditions. Advection periods
longer than 30 minutes with constant upwind blowing directly from a large emission source
cannot be removed by the stationary test. To investigate this latter effect, we plotted the one
129
minute average CO2 concentrations measured during the entire study against the mean wind
velocity and wind direction trying to identify build-ups of elevated concentrations related to
specific upwind characteristics. We found that only four periods from nine possible periods with
advection were not removed by the stationarity test. The high CO2 concentrations observed in
these four periods were not because of daily activities, and although they present fluxes above
the average at that time of the day, they do not represent spikes and their influence in the diurnal
course is not significant. Therefore, we consider these periods as part of the variability in our
results.
4.3.6. Footprint analysis
In a homogeneous surface, the footprint is not an issue because the fluxes from all parts of
the surface are by definition equal. However, in an inhomogeneous surface such as a city, the
measured signal depends on which part of the surface has the strongest influence on the sensor
and, therefore, the location and size of the footprint. To evaluate the footprint in our experiment,
we used the model developed by Hsieh et al. (2000) with the previously estimated zero plane
displacement and an assumed roughness height of 1 m. We applied this model to the complete
set of 30-minute periods measured during the campaign to determine the fraction of the flux
measured (F/S0) as a function of the upwind distance (x) and the atmospheric stability condition.
F represents the flux and S0 the source strength. If the footprint is defined to encompass 80% of
the total flux, we observe in Figure 4 that the longest footprints (~ 5 km) correspond to stable
conditions, which prevail at nighttime; while the smaller footprints (500 m) correspond to
unstable conditions, characteristic of morning and early afternoon. The average footprint,
considering all stability conditions, was 1.3 km, which represents a distance large enough to
characterize the CO2 fluxes of a typical neighborhood in Mexico City. Figure 5 shows different
footprint fractions for selected time periods of the day. At night, the footprint is distributed over
larger distances; 80% of its accumulation occurs in a radius between 1.5 and 4 km. The unstable
conditions during morning and early afternoon produce shorter footprints, and from a
130
predominant direction between south and southeast. During late afternoon, unstable conditions
also prevail, but with less frequency and strength, to produce footprints longer than in previous
hours and smaller than during the night.
4.4. Results
4.4.1. Concentrations
The CO2 concentrations showed a clear diurnal pattern similar to other pollutants emitted by
mobile sources, such as NOX and CO (INE, 2000). As indicated in Figure 6, the highest
concentrations occurred between 6:30 and 8:00 h (all times are reported in LST) with a range
from 398 to 444 ppm and an average of 421 ppm. This morning peak is attributed to
anthropogenic emissions (mainly vehicles), nocturnal respiration, and the shallow early morning
mixed layer. The growth of the boundary layer depth for Mexico City has been identified to
occur between 10:00 and 12:30 h (Whiteman et al., 2000), passing from mixing heights lower
than 500 m to heights above 2000 m at noon. The lowest CO2 concentrations were observed
during the afternoon, with an average of 375 ppm. In the afternoon, the growth of the boundary
layer slows significantly, and begins to decay around 18:00 h, when the CO2 concentration starts
rising again. This CO2 concentration diurnal profile is consistent with observations at other
urban sites, such as Chicago (Grimmond et al., 2002). The diurnal pattern was relatively
constant during the entire study. However, an average difference of 20 ppm was observed
during the morning rush hour between the Holy Week (week 2) and the first week of the
campaign. This difference represents the vehicular traffic reduction due to the national holiday
period during week 2. The average difference between week 1 and week 3, in which almost all
schools were still on holiday was 6 ppm. The mean diurnal concentration of 388 ppb with a
standard deviation of 14 ppm observed in this study was similar to those reported for other cities
(Grimmond et al., 2002; Nemitz et al., 2002; Reid and Steyn, 1997) and also to measurements
collected in Mexico City in 2001 (Grutter, 2003). However, Grutter identified a later and longer
morning rush hour peak in CO2 concentrations, probably due to a more dense network of roads
131
near his site, and a second peak at night that we did not observe. The nighttime peak in Grutter’s
measurements was probably because his measurements were made next to a large forest reserve.
4.4.2. Fluxes
The CO2 flux measurements showed a clear diurnal pattern, with the highest emissions
during the morning, and lower emissions during nighttime (see Figure 7). Fluxes ranged from -
0.22 to 1.60 mg m-2 s-1, with an average of 0.41 mg m-2 s-1. Although negative fluxes were
observed occasionally during unstable conditions, fluxes were predominately positive during our
study. This indicates that the urban surface is a net source of CO2, and that the uptake by urban
vegetation during the day is not strong enough to offset the flux from anthropogenic sources.
Fluxes exhibited a relatively constant diurnal pattern during the three study weeks. The
highest fluxes were observed in the first week, and the lowest during the second week. Figure 7
shows the average diurnal patterns for weekdays and weekends. Lower fluxes occurred on
weekends between 5:00 and 9:00 h compared to fluxes at these times on weekdays. This effect
is directly related to vehicular traffic, since on weekends the traffic is usually lower during
morning hours. During the remainder of the day, fluxes are similar between weekends and
weekdays.
Diurnal patterns of concentrations and fluxes are different. Concentrations not only reflect
the variability in the emissions, but also the evolution of the atmospheric boundary layer. During
the early morning, concentrations increase until 7:00 h, while fluxes remain constant until 6:00 h,
when they start to increase, reaching their maximum at 9:00 h. At this hour, concentrations have
already started to decline, continuing until noon, while fluxes remain high during the entire
morning. This shows the dilution of CO2 from the actual emissions and from the pool stored in
the residual layer, as a function of the atmospheric boundary layer growth. Both concentrations
and fluxes remain constant during the afternoon. After 19:00 h, fluxes start to decrease, but
concentrations increase again, indicating that a new nocturnal boundary layer has started to form.
132
The CO2 flux diurnal pattern in Figure 7 shows three consistent peaks during the morning,
7:00, 9:00 and 11:30 h. The earliest peak was observed only during weekdays, while the other
two peaks were observed on both weekdays and weekends. The peak at 7:00 h was observed in
20% of the days, the peak at 9:00 h 40% and the peak at 11:30 h in 35% of the cases. These
peaks are not related to the upwind direction or any other measured parameter and they do not
appear to be caused by horizontal advection as discussed previously. Overall our results show
that the CO2 fluxes are higher during the morning than during the afternoon, and that the
morning peaks illustrate the variability of the morning CO2 emissions in the study area. With a
longer campaign, the features involved with those peaks might be resolved, or they could be
absorbed in a broad morning peak.
Overall, the magnitude and diurnal pattern of the CO2 fluxes in Mexico City are between
those observed in winter and summer in mid-latitude cities. In mid-latitude cities, the CO2 fluxes
depend on the season. During summer CO2 emissions tend to be lower, the benign weather
reduces the fuel consumption for heating and trees are in full leaf increasing the CO2 uptake by
photosynthesis. During winter the opposite occurs, CO2 emissions are accentuated by heating
necessities and CO2 uptake by vegetation is reduced. For example, reported CO2 fluxes in
Tokyo during July vary between 0.2 and 0.5 mg m-2 s-1, while during December measured fluxes
in Tokyo are between 0.2 and 1.1 mg m-2 s-1 (Moriwaki and Kanda, 2004). Our CO2 fluxes
match both the magnitude and diurnal pattern of fluxes reported by Walsh et al. (2004) for a
residential area of Vancouver for April and June.
In addition to the CO2 fluxes, sensible and latent heat fluxes were also measured with the
same flux system. Sensible heat flux was always directed upward with peak values around 270
W m-2 at noon, while latent heat fluxes were small with peak values below 90 W m-2 also at
midday. From heat fluxes reported previously for other urbanized sections of Mexico City
during the cold season (Oke et al., 1999; Tejeda and Jauregui, 2005), we found that differences
in the energy balance between the coldest and warmest months are not significant, and if we add
133
that the scarce vegetation cover in Mexico City is relatively constant throughout the year, we can
infer that CO2 fluxes should be also relatively constant throughout the annual cycle.
4.4.2.1. Fluxes as a function of the upwind direction
In an urban area, where the emission sources are not homogeneous, it is necessary to
determine the magnitude of the fluxes as a function of the upwind direction to identify the
strongest sources. Figure 8 shows the CO2 flux distribution as a function of the wind direction
superimposed on a map of the study area. Fluxes were lower from areas with fewer streets, for
example from the north and northwest directions. The largest emissions came from the east and
southeast directions, where the avenue (Anillo Periferico) with the highest vehicular traffic near
the measurement site is located. The vegetation cover in the National park “Cerro de la
Estrella”, located southwest from the tower, should be a sink for CO2 emitted in the region, but it
does not appear to be strong enough to offset local emissions.
4.4.2.2. Fluxes as a function of the vehicular activity
During the MCMA field campaign, traffic counts were performed at two intersections, one
between the avenues labeled 1 and 2 (I1-2) in Figure 8, and the other between avenues 2 and 3 (I2-
3). Figure 9 presents the diurnal flux pattern for the 45º upwind sectors corresponding to each
intersection along with the vehicular count distributions, where the vehicular fleet was classified
into four different groups: passenger cars, taxi cabs, light trucks/buses, and heavy trucks/buses.
Although both intersections share avenue 2, I2-3 has a higher traffic density than I1-2, since
avenue 3 encircles the entire city, making it one of the busiest avenues, and one of the largest
emission sources. Several routes of “colectivos” (small buses with a 20 passenger capacity) start
from I1-2, which can be observed in the higher number of light buses and trucks compared to I2-3.
The high CO2 fluxes emitted from the southwest shown in Figure 8 may be due to these
colectivos, which remain stopped and idling at a short distance from this intersection, while
waiting for passengers.
134
The typical morning and afternoon traffic peaks are not identified at either intersection,
instead a single peak is observed during the entire day. This peak begins at 6:00 h, and extends
until late afternoon for I1-2, and into the night for I2-3. The morning and noon gaps in the CO2
flux courses for the upwind sectors corresponding to I1-2 and I2-3 are because of the lack of
sampling periods with winds from those directions. Although two large peaks are observed in
the CO2 fluxes (16:30 h at I1-22 and 9:30 h at I2-3), we conclude that CO2 fluxes follow the
vehicular traffic diurnal profile. Both remain constant during most of the day and then decline at
night. Figure 10 confirms the high correlation between vehicular traffic and CO2 emissions,
where the correlation coefficients from linear regressions are 0.85 and 0.58 for I1-2 and I2-3,
respectively. The higher correlation coefficient for I1-2 is because that intersection was closer to
the flux tower (1580 m) than the intersection I2-3 (2100 m). The statistical offset indicates the
presence of other sources, such as cooking, water heating, and combustion processes from
factories and stores, among others. According to the greenhouse gas emission inventory for
Mexico City, the transportation sector contributes with 60% to the CO2 emissions burden (33.6
Mton year-1 for the year 1996), followed by the industrial and residential sectors, with
contributions of 20% and 16%, respectively (Sheinbaum et al., 2000).
Using the Mexican vehicular fleet distribution (SMA-GDF, 2002), the emission factors
developed by Zavala (2004) using the stoichiometric combustion equation, and the
characteristics of Mexican fuels, we roughly determined the fraction of CO2 emitted by each
vehicle class for the two intersections discussed above. We found that the emission
contributions are similar for both intersections, where the group of passenger cars has the largest
vehicular source (60%). Colectivos and taxis contribute 17% each, while heavy trucks and buses
represent 6% of the mobile CO2 emissions.
4.4.2.3. Evaluation of random and systematic errors in the measured daily mean CO2 flux
Although error propagation cannot be quantified precisely for flux measurements, a
sensitivity analysis on the effects of errors on an average day basis can help to quantify the flux
135
uncertainties. We applied the approach proposed by Moncrieff et al. (1996) to evaluate the
random and systematic errors on the mean daily flux. When errors are random, errors in
computed means diminish with increasing size of data set according to N-1/2, where N is the
number of measurements. Thus random errors can be estimated by examining the convergence
of calculations of the net flux. In contrast, systematic errors are not affected by increasing data
set size, because they simply add in a linear fashion. In general, systematic errors are difficult to
detect.
Figure 11a shows the results of applying diverse values of a percentage random error (pr) to
each half hour CO2 flux in the entire data set. The horizontal line represents the mean daily flux,
0.41 mg m-2 s-1. The symmetrical set of lines above and below this line represent the overall
random error as a function of the magnitude of pr and the given number of days. If we assume a
random error of +20% for each 30-minute mean flux for the 23 days studied, the total random
error on the net flux is 0.202 mg m-2 s-1 (50% of the net flux). If we also assume a systematic
error (ps) of +10%, we see from Figure 11b that the total systematic error on the net flux is 0.041
mg m-2 s-1. Adding these two errors, the total uncertainty of the net flux is 0.243 mg m-2 s-1 (60%
of the total flux). This technique allows us to determine the required number of days to obtain
the best possible flux estimate. For example, considering a random error of 20% and a
systematic error of 10%, 100 days of measurements would resolve a net flux of 0.508 mg m-2 s-1,
25% higher than our observed mean daily flux. One year of measurements would resolve a net
flux of 0.460 mg m-2 s-1 and reduce the offset to 13%. However, flux measurements can only be
compiled for a certain length of time before possible seasonal trends become important.
Although the weather in Mexico City throughout the annual cycle is relatively constant in
comparison to the weather in mid-latitude cities, three climatic seasons can be differentiated, a
dry, warm season from March to May, a rainy season from June to October, and a dry, cold
season from November to February. Thus, a seasonal evaluation of CO2 fluxes in Mexico City
will require measurement periods of 3 months for each season; for this duration the estimated
error would be 36% based on a random error of 20% and a systematic error of 10%.
136
4.5. Conclusions
Results of this study have demonstrated the capacity of the eddy covariance technique to
measure CO2 fluxes in an urban area, where the inhomogeneous surface, high roughness, and
wide distribution of emission sources complicate the measurements. Micrometeorological
techniques are capable of quantifying CO2 emissions from sources that are not usually accounted
with other methods, such as factories hidden in residential areas or street food cooking.
Our CO2 measurements show clear diurnal patterns for both concentrations and fluxes, which
are strongly correlated to vehicular traffic during the day. The consistently positive CO2 fluxes
measured in Mexico City are in agreement with other urban studies, which have shown that the
urban surface is a net source of CO2 (Grimmond et al., 2002; Nemitz et al. 2002; Walsh et al.,
2004 ; Moriwaki and Kanda et al., 2004; Grimmond et al., 2004). The effects of CO2 uptake by
urban vegetation during the daytime do not appear to be significant for this site.
Even though the weather and vegetation cover in Mexico City is relatively constant, some
differences in the CO2 fluxes can be expected during the different climatic seasons of the year.
But if we assume that these differences are small, and that fluxes measured during the studied
period could be representative of the entire year, we could infer from the average flux an annual
CO2 emission of 12.8 kg m-2. Annual emissions obtained from other cities using similar flux
techniques agree quite well with our measurements. Moriwaki and Kanda (2004) reported an
annual CO2 emission 4% lower than Mexico City for a residential area of Tokyo, while Soegaard
and Møller-Jensen (2003) determined essentially the same annual emission for the urban area of
Copenhagen. This finding is somewhat surprising, since we expected to observe higher
emissions in Mexico City compared to more developed cities, because of the higher population,
scarce green areas, dense traffic with a relatively old vehicle fleet, and the extensive presence of
aged technologies for daily commercial and residential activities. In Mexico City, CO2
emissions from combustion sources are accentuated by the high elevation of the city, which
makes the combustion process less efficient.
137
The ability to make CO2 flux measurements over an urban landscape is an important step in
improving our understanding of the role megacities play in global climate change. Because
emission inventories for CO2 are typically based on fuel inventories or by other indirect means
and since this kind of information may not be available or reliable in less developed areas, it is
important to evaluate available emission estimates with direct measurements. In the future, long
term measurements similar to those described in this paper would be a valuable contribution to
quantification of CO2 emissions from megacities. Fluxes reported here are for a short period of
time; clearly more measurements are needed to confirm the representativeness of the CO2 fluxes
measured in Iztapalapa for a poor neighborhood of Mexico City and their temporal variability
throughout the year. In a separate paper (Velasco et al., 2005), we have also demonstrated that
these flux methods can also be applied to measure VOCs that are important for urban and
regional atmospheric chemistry.
4.6. Acknowledgements
This study was supported by the Integrated Program on Urban, Regional and Global Air
Pollution from the Massachusetts Institute of Technology (MIT) in collaboration with the
Metropolitan Commission of Environment (CAM) of Mexico City. The authors would like to
acknowledge the support and assistance of Mario and Luisa Molina as the directors of the
MCMA-2003 field campaign, as well as the National Center for Environmental Research and
Training (CENICA) of the Mexican National Institute of Ecology (INE) as hosts of this study.
Authors also wish to thank the two anonymous reviewers, who provided very helpful comments
for the final manuscript.
138
4.7. References
Aubinet, M., Grelle, A., Ibrom, A., Rannik, Ü., Moncrieff, J., Foken, T., Kowalsky, A.S., Martin, P.H., Berbigier, P., Bernhofer, Ch., Clement, R., Elbers, J., Granier, A., Grünwald, T., Morgenstern, K., Pilegaard, K., Rebmann, C., Snijders, W., Valentini, R., Vesela, T., 2000. Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology. Advances on Ecological Research 30, 113-175.
Auble, D.L., Meyers, T.P., 1992. An open path, fast response infrared absorption gas analyzer for H2O and CO2. Boundary Layer Meteorology 59, 243-256.
Baez, A., Reyes, M., Rosas, I., Mosiño, P., 1988. CO2 concentrations in the highly polluted atmosphere of Mexico City. Atmósfera 1, 87-98.
Baldocchi, D., Falge, E., Gu, L., Olson, R., Hollinger, D., Running, S., Anthoni, P., et al., 2001. FLUXNET: a new tool to study the temporal and spatial variability of ecosystem scale carbon dioxide, water vapor and energy flux densities. Bulletin of the American Meteorological Society 82, 2414-2434.
Grimmond, C.S.B., Salmond, J.A., Oke, T.R., Offerle, B., Lemonsu, A., 2004. Flux and turbulence measurements at a densely built-up site in Marseille: heat, mass (water and carbon dioxide), and momentum. Journal of Geophysical Research 109, 24101-24120.
Grimmond, C.S.B., King, T.S., Cropley, F.D., Nowak, D.J., Souch, C., 2002. Local-Scale fluxes of carbon dioxide in urban environments: Methodological challenges and results from Chicago. Environmental Pollution 116, 243–254.
Grimmond, C.S.B., Oke, T.R., 1999. Aerodynamic properties of urban areas derived from analysis of surface form. Journal of Applied Meteorology 38, 1262-1292.
Grutter, M., 2003. Multi-gas analysis of ambient air using FTIR spectroscopy over Mexico City. Atmósfera 16, 1-13.
Hsieh, C.I., Katul, G. & Chi, T., 2000. An approximate analytical model for footprint estimation of scalar fluxes in thermally stratified atmospheric flows. Advances in Water Resources 23, 765-772.
INE, 2000. Almanaque de datos y tendencias de la calidad del aire en ciudades mexicanas. Instituto Nacional de Ecología, SEMARNAT. México D.F.
INEGI, 2000. Estados Unidos Mexicanos Resultados Prelimanares XII Censo General de Población y Vivienda. INEGI, México D.F.
Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric boundary layer flows. Their structure and measurement. Oxford University Press, US.
139
Massman, W.J., Lee, X., 2002. Eddy covariance flux corrections and uncertainties in long-term studies of carbon and energy exchanges. Agricultural and Forest Meteorology 113, 121-144.
McMillen, R., 1988. An eddy correlation technique with extended applicability to non-simple terrain. Boundary Layer Meteorology 43, 231-245.
Moncrieff, J.B., Malhi, Y., Leuning, R., 1996. The propagation of errors in long-term measurements of land-atmosphere fluxes of carbon and water. Global Change Biology 2, 231-240.
Moriwaki, R., Kanda, M., 2004. Seasonal and diurnal fluxes of radiation, heat, water vapor, and carbon dioxide over a suburban area. Journal of Applied Meteorology 43, 1700-1710.
Nemitz, E., Hargreaves, K.J., Mcdonald, A.G., Dorsey, J.R., Fowler, D., 2002. Micrometeorological measurements of the urban heat budget and CO2 emissions on a city scale. Environmental Science and Technology 36(14), 3139-3146.
Oke, T.R., Spronken-Smith, R.A., Jauregui, E., Grimmond, C.S.B., 1999. The energy balance of central Mexico City during the dry season. Atmospheric Environment 33, 3919-3930.
Reid, K.H., Steyn, D.G., 1997. Diurnal variations of boundary-layer carbon dioxide in a coastal city–observations and comparison with model results. Atmospheric Environment 31, 3101-3114.
Sheinbaum, C., Ozawa, L.V., Vasquez, O., Robles, G., 2000. Inventario de emisiones de gases de efecto invernadero asociados a la producción y uso de energía en la Zona Metropolitana del Valle de México. Reporte Final del Grupo de Energía y Ambiente. Universidad Nacional Autónoma de México, México D.F.
Schmid, H.P., Grimmond, C.S.B., Cropley, F., Offerle, B., Su, H.B., 2000. Measurements of CO2 and energy fluxes over a mixed hardwood forest in the mid-western United States. Agricultural and Forest Meteorology 103, 357-374.
Soegaard, H., Møller-Jensen, L., 2003. Towards a spatial CO2 budget of a metropolitan region based on textural image classification and flux measurements. Remote Sensing of Environment 87, 283-294.
SMA-GDF, 2004. Inventario de Áreas Verdes en el Distrito Federal. Gobierno del Distrito Federal, México D.F. Available in the website: http://www.sma.df.gob.mx. Visited on May 2005.
SMA-GDF, 2002. Inventario de emisiones Zona Metropolitana del Valle de México 1998. Gobierno del Distrito Federal, México D.F.
Tejeda, A., Jáuregui, E., 2005. Surface energy balance measurements in the Mexico City region: a review. Atmósfera 18, 1-23.
140
Velasco, E., Lamb, B., Pressley, S., Allwine, E., Westberg, H., Jobson, T., Alexander, M., Prazeller, P., Molina, L., Molina, M., 2005. Measurements of urban VOC fluxes. Geophysical Research Letters 32, doi: 10.1029/2005GL023356.
Vogt, R., Christen, A., Rotach, M.W., Satyanarayana, A.N.V., 2003. Fluxes and profiles of CO2 in the urban roughness sublayer. Fifth International Conference on Urban Climate. September 1-5, 2003, Lodz, Poland.
Walsh, C.J., Oke, T.R., Grimmond, C.S.B., Salmond, J.A., 2004. Fluxes of atmospheric carbon dioxide over a suburban area of Vancouver. Fifth Symposium on the Urban Environment 23 - 27 August, 2004. Vancouver, Canada.
Webb, E.K., Pearman, G.I., Leuning, R., 1980. Correction of flux measurements for density effects due to heat and water vapour transfer. Quarterly Journal of the Royal Meteorological Society 106, 85-100.
Westberg, H., Lamb, B., Hafer, R., Hills, A., Shepson, P., Vogel, C., 2001. Measurement of isoprene fluxes at the PROPHET site. Journal of Geophysical Research 106(D20), 24,347-24,358.
Whiteman, C.D., Zhong, S., Bian, X., Fast, J.D., Doran, J.C., 2000. Boundary layer evolution and regional-scale diurnal circulations over the Mexico Basin and Mexican plateau. Journal of Geophysical Research 106(D8), 10,081-10,102.
Zavala, M., 2004. Personal communication. Massachusetts Institute of Technology, Boston, MA.
141
Figure 4.1. Schematic diagram of the instrumented flux tower. The azimuth orientation of the
sonic anemometer is 16º from north. Dimensions are in meters.
142
Figure 4.2. (a) Power density spectra for CO2 concentration and ambient temperature,
normalized for comparison. The -5/3 slope indicates the theoretical slope in the inertial
subrange. (b) Cospectra of vertical velocity with ambient temperature and CO2 concentration,
normalized for comparison. X represents the average CO2 concentration and ambient
temperature for the spectra, and the covariances of those scalars with the vertical wind speed for
the cospectra. Overall, the shape and details of the CO2 and ambient temperature spectra and
cospectra correspond closely over the entire range of frequencies measured. The data correspond
to the 30 minute sampling period on April 7 at 15:30 h.
143
Figure 4.3. Stationarity test for CO2 flux: in 56% of the periods, the flux difference was less
than 30%, which indicates periods that meet and exceed the stationarity criteria. In 18% of the
periods, the flux difference was between 30 and 60%, which means that these periods have an
acceptable quality.
144
Figure 4.4. Fraction of the flux measured (F/S0) versus the upwind distance or effective fetch
(x).
145
Figure 4.5. Different fractions of the measured flux (F/S0) during the entire campaign as
function of the wind direction for different intervals of time, (a) from 0:00 to 3:00 h, (b) from
6:00 to 9:00 h, (c) from 12:00 to 15:00 and (d) from 18:00 to 21:00 h.
146
Figure 4.6. Average diurnal pattern of CO2 concentrations for the entire study and for separate
weeks. The gray shadow represents ±1 standard deviation from the total average.
147
Figure 4.7. Average diurnal pattern of CO2 fluxes for the entire study and for weekdays and
weekends. The gray shadow represents ±1 standard deviation from the total average.
148
Figure 4.8. CO2 flux distribution as a function of the upwind direction during the entire study.
The contour shape indicates the fraction of the flux measured equal to 80% and the shading
indicates the magnitude of the CO2 fluxes. The black spot indicates the position of the flux
tower; and the solid area in the left bottom corner represents part of the National park “Cerro de
la Estrella”. The four primary roads surrounding the measurement site are: (1) Av. Rojo
Gomez, (2) Av. Ermita Iztapalapa, (3) Anillo Periferico, (4) Av. Jalisco, and (5) Calz. San
Lorenzo.
149
Figure 4.9. Diurnal profile of CO2 fluxes superimposed over plots of traffic counts for two
intersections within the footprint: intersection I1-2 between Av. Ermita Iztapalapa and Av. Rojo
Gomez, and intersection I2-3 between Av. Ermita Iztapalapa and Anillo Periferico. Fluxes
correspond to the upwind sectors where both intersections are located, for I1-2 the sector between
240 and 285º and for I2-3 the sector between 115 and 160º.
150
Figure 4.10. Correlation between CO2 fluxes and vehicular traffic for two intersections, I1-2 and
I2-3. Fluxes correspond to the 45º upwind sectors where both intersections are located,
respectively.
151
Figure 4.11. Effects of random and systematic errors for the mean daily CO2 flux. (a) Effects
due to random errors for various percentages of error. (b) Effects due to systematic errors as a
function of the error percentage (ps). In both figures the horizontal line represents the mean daily
CO2 flux, 0.41 mg m-2 s-1.
152
153
CHAPTER 5
SUMMARY AND CONCLUSIONS
5.1. Summary
The work presented in this dissertation investigated the ambient concentrations and fluxes of
volatile organic compounds (VOCs) and CO2 in the atmosphere of Mexico City. There are many
individual reactive VOCs in the troposphere that contribute to the creation of photochemical
smog and tropospheric ozone. In addition, some individual VOCs are toxics, e.g., the
carcinogenic benzene and 1-3-butadiene. CO2 is the dominant greenhouse gas of anthropogenic
origin.
The overall goal of this research was to quantify the emissions of these trace gases in a
megacity. Better emissions should lead to better strategies for reducing air pollution. To pursue
this goal, we participated in the Mexico City Metropolitan Area (MCMA) field campaigns in
2002 and 2003. Our participation involved two main activities. First we investigated the ambient
concentrations of VOCs in terms of their distribution, diurnal pattern, origin and reactivity at
different locations of the Valley of Mexico. This task was completed in collaboration with other
participants in the MCMA field campaigns, who provided us with VOC data. Our second
activity involved the deployment of micrometeorological flux instruments on a tall tower in a
typical Mexican neighborhood to directly measure fluxes of selected VOCs and CO2 from the
urban landscape. Fluxes of olefinic VOCs were measured using a Fast Olefin Sensor (FOS) and
processed by the eddy covariance technique (EC). Fluxes of acetone, methanol, C2-benzenes and
toluene were measured using a Proton Reaction Mass Spectrometer (PTR-MS) and computed by
the disjunct eddy covariance technique (DEC). Fluxes of CO2 were measured by an open path
infrared sensor and processed by EC.
The use of different techniques to measure ambient VOC concentrations provided
confidence in the data, as well as a basis for comparison with grid model simulations of selected
VOCs. During the MCMA-2002 field campaign only canister samples were collected and
154
analyzed by gas chromatography/flame ionization detection (GC-FID) with the aim to determine
the VOC signature at different sites within and near Mexico City. Those monitoring sites
included locations in the urban core, a heavily industrial area and boundary sites in rural
environments. For the MCMA-2003 field campaign the VOC measurements were extended in
both, monitoring sites and instrumentation. Besides of GC-FID, VOC concentrations were
measured by two PTR-MS, one FOS and, two UV Differential Optical Absorption Spectrometers
(DOAS) located in different sites of the city. The following points summarize the main findings
of the ambient VOC measurements:
- During the morning the entire Valley of Mexico presents a homogeneous distribution of VOCs:
60% alkanes, 15% aromatics, 5% olefins and a remaining 20% of oxygenated, alkynes, esters
and other unidentified VOCs (percentages are based on concentrations in pbbC). In the
afternoon, the VOC mix changes according to the characteristics of each site. If reactivity with
OH and atmospheric abundance are used to measure photochemical reactivity, olefins are the
most important hydrocarbons in the atmosphere of Mexico City. The top three contributors are
ethylene, propylene and propane. Even though propane and butane are not overly reactive with
OH, their elevated concentrations in Mexico City are sufficient to rank these two alkanes in the
top 5 reactive VOCs.
- Elevated levels of toxic hydrocarbons, such as 1,3-butadiene, vinyl chloride and the BTEX
hydrocarbons were observed. These hydrocarbons are of significant public health concern since
there is evidence that they increase the risk of cancer, leukemia, congenital malformations and
many other adverse health effects.
- In general, VOC mixing ratios observed in this study were slightly lower than those reported in
earlier studies in Mexico City. This observation is consistent with previous reports indicating
that ambient VOC concentrations have reach a stabilization level and possibly have decreased.
This is a good indicator that policies and actions enacted to decrease VOC emissions have been
successful, despite the growth in the vehicular fleet and other activities.
155
- VOC diurnal patterns show a strong correlation with the daily vehicular traffic pattern. At
some urban sites, one single peak concentration is observed coinciding with the morning rush
hour and the major vehicular activity of the day. In contrast, other sites show two peaks
corresponding to the morning and afternoon rush hours, when increments in the vehicular
activity are also registered. Diurnal patterns at boundary sites are not influenced by vehicular
traffic or biogenic emissions; instead they depend more on wind patterns and rural sources such
as burnings of agriculture debris and trash.
- Comparisons of ambient VOC concentrations with concentrations of acetylene and VOCs with
similar lifetime demonstrated that many olefin and aromatic species, such as xylenes, toluene,
ethylbenzene, t-2-pentene, c-2-pentene, etc. are mainly emitted in vehicle exhausts,. Also,
MTBE and some alkanes such as 2-methylpentane and 3-methylpentane showed excellent
correlations with acetylene, indicating that their main source is also vehicle exhaust.
Furthermore, comparisons of light alkanes with acetylene showed that certain species related
with the use of residential LPG are emitted also by mobile sources, particularly propane and n-
butane.
- Examination of the VOC data in terms of lumped modeling VOC classes and comparison to the
local emissions inventory suggests that some, but not all, classes are underestimated in the
inventory by factors between 1 and 3. Overall, the emissions inventory underestimates the
contribution of some alkanes and overestimates the contributions of some olefins and aromatics.
The flux measurements presented in this dissertation represented the first experiment in
which micrometeorological techniques are applied to measure fluxes of VOCs in an urban
environment. Micrometeorological techniques have not previously been used in urban areas to
measure VOC fluxes because of the lack of fast-response VOC sensors and the uncertainties
associated with flux measurements in a city, where the spatial variability of surface cover,
emission sources and roughness is high. However, in this study, we have demonstrated the use
of EC and DEC techniques to perform VOC flux measurements in an urban area using state of
156
the art VOC sensors. The CO2 EC measurements represented the first CO2 fluxes quantified in
an urban area of a developing country. The most significant findings from our flux
measurements are listed below.
- Fluxes of olefins and CO2 exhibited a clear diurnal pattern with a strong relationship to
vehicular traffic. In addition, examination of these fluxes as a function of the upwind direction
indicated that the highest fluxes were observed in sectors with a high density of streets and
avenues with heavy traffic.
- Olefin fluxes were lower on the weekends than on weekdays, and on weekends the morning
rush hour peak was not observed. On average, the daily flux on weekdays was 21% higher than
during weekends. This effect is directly related to changes in vehicular traffic and to industrial
and commercial activities.
- Fluxes of methanol, acetone, toluene and C2-benzenes showed similar diurnal patterns, with the
highest fluxes during the morning and lowest during night. In general, fluxes for these species
were always positive. Fluxes of acetone were negative in early morning, which suggests that
deposition of acetone may occur before sunrise.
- The CO2 flux measurements showed that the urban landscape is a net source for CO2 with
similar diurnal patterns and magnitudes to those observed in US or European cities. The effects
of CO2 uptake by urban vegetation do not appear to be significant, at least for the analyzed
footprint in this study.
- The diurnal profile of olefins, acetone, toluene and C2-benzenes fluxes was compared with the
emission profiles reported in the local gridded emissions inventory. There was general
agreement between measured fluxes and reported emissions, with the exception of C2-benzenes,
whose emissions in the inventory were more than two times higher than the measured fluxes.
157
5.2. Conclusions
In this dissertation indirect and direct methods to evaluate emissions inventories were applied
to the actual emissions inventory of VOCs for Mexico City. Detailed measurements of
individual VOCs and CO2 concentrations were also conducted with the aim to better understand
their distribution, magnitudes, diurnal patterns, and principal emission sources. Therefore, from
all our evaluations and analysis we derive the following conclusions:
- In this work, we have demonstrated the use of EC and DEC techniques to perform VOC flux
measurements in an urban area using state of the art VOC sensors. The ability to evaluate
emissions inventories using micrometeorological techniques is a valuable and new tool for
improving air quality management.
- Comparisons of the measured fluxes versus the emissions of olefin, toluene and acetone
reported in the most recent emissions inventory for air quality modeling indicate that fluxes and
emissions agree for the grid where the flux tower was located and do not support recent
modeling results suggesting that VOC emissions are underestimated by a factor of 3. If this is
true for other VOC classes and in other sections of the city, it implies that factors other than an
underestimation of VOC emissions may be causing poor agreement between modelled and
measured ozone levels. At the same time, it is important to note that the species we measured do
not account for a majority of the VOC emissions which tend to be dominated by alkane species.
For this reason, it is important to extend these direct VOC flux measurements to include other
compounds and to also make measurements in other parts of the city. The comparison between
the ambient VOC concentrations in terms of lumped modeling classes versus the emissions
inventory neither support the statement that VOC emissions are underestimated by a factor of 3.
- We found that the developing megacity of Mexico City emits essentially the same annual
amount of CO2 per unit of area as that released in developed cities such as Tokyo and
Copenhagen. This is in spite of the higher population, scarce green areas, dense traffic with a
relatively old vehicle fleet, and the extensive presence of aged technologies for daily commercial
158
and residential activities in Mexico City. The higher fuel consumption in developed cities could
explain this similitude.
- Through the different analysis performed in this work, vehicle exhaust was identified as the
main source of VOCs and CO2 in the atmosphere of Mexico City. This result suggests that
control strategies need to be focused on improving fuel quality and vehicle emission technology.
However, these kinds of control strategies alone are not going to solve the air pollution problem
of a developing megacity, where not all of the population have access to cleaner transportation
technologies and less polluting methods for cooking and water heating. First, the economic
necessities of the major population need to be solved, and the actual style of life needs to be
modified. It is not sustainable to have private cars being the main transport mode. New kinds of
high-quality mass public transport need to be implemented to reduce the use of private cars.
Also, new policies promoting environmental education, the use of the bicycle and a greater
support to environmental research are needed to reduce the harmful VOC and CO2 levels
reported here, and overall, the air pollution problem in Mexico City.
5.3. Advantages and disadvantages of a flux system in an urban area
As an additional conclusion based on the experience gained through this study, we present a
list of the advantages and disadvantages of a flux system in an urban area to evaluate fluxes of
trace gases.
Advantages:
- Direct flux measurements during the entire diurnal cycle.
- Experimental validation of gridded emission inventories.
- Footprint of 0.5 to 5 km (significant fraction of one model grid).
- Simultaneous measurements of energy fluxes and turbulence data for evaluation of urban
boundary layer models.
159
- Micrometeorological techniques are capable of quantifying emissions from sources that can not
be identified, such as factories hidden in residential areas or street food cooking.
Disadvantages:
- One fixed site.
- Uncertainties about sensor sensitivity and source strengths.
- Effects of chemical loss on measured fluxes.
- Uncertainty about footprint in urban area.
- The cost of the required fast response sensors may be high.
5.4. Future work
In order to reduce the disadvantages listed above and to solve the uncertainties marked in this
dissertation about flux systems in urban environments, it is clear that more research is needed. In
this context we plan to go back to Mexico City and collect VOC flux measurements at a new site
as part of a large field campaign named “Megacity Initiative: Local and Global Research
Observations” (MILAGRO) scheduled for March 2006. In this second flux study we expect to
answer the question whether the results from our MCMA-2003 campaign would hold for other
locations in the city and for other VOC compounds. Answers to these questions will be
important for modeling analyses and for future air quality management within Mexico City and
other megacities.
We will identify a suitable location for flux measurements where the emissions are relatively
high and where a tower can be safely erected and instrumented. We have already located the
building hosting the offices of the local atmospheric monitoring network (RAMA) situated near
central Mexico City where traffic densities are high and there is a range of residential and
commercial facilities. Again we will measure olefin and CO2 fluxes using a FOS and IRGA
sensor, respectively, and the number of aromatic and oxygenated VOCs measured by PTR-MS
160
will be expanded. Also, we plan to expand the list of all VOC compounds measured by using a
disjunct eddy accumulation (DEA) method coupled with individual VOC GC-FID analyses.
In the DEA method, updrafts and downdrafts are detected with the sonic anemometer and air
is sampled during the up (down) draft into an up (down) sample volume at a volume proportional
to the strength of the up (down) draft. The flux is calculated from measured velocities and
concentrations using:
WCCwCwF −+= −−++ (1)
where w+ and w- are the mean updraft and downdraft velocities, C+ and C- are the corresponding
mean concentrations, and W and C are the overall means during the sampling period. With this
system, the up/down samples are collected during a 30 min period and then require analyses
using a gas chromatography system. With this batch approach, the frequency of sampling is
limited to approximately 2 hours. Thus, we plan to collect 30 min flux samples on a 2 hour basis
during each sample day. For some days, the frequency can be increased, for example every hour
during the morning rush hour, to obtain more detailed data during selected periods. This system
will give us information on compounds not measured otherwise and also provide data for some
species for intercomparison to the other flux methods.
In addition, we plan to measure fluxes of CO. Emissions of VOCs and CO2 from combustion
sources are strongly correlated to emissions of CO. CO fluxes will valuable to identify better the
emission sources of VOCs and CO2.
Prior to the MILAGRO campaign we plan to reduce the uncertainty of the 52% of olefins
detected by the FOS that remain unknown through response tests. And after the campaign,
through comparison of FOS data with the individual VOC analyses from the DEA flux system.
Beyond the MILAGRO campaign, we suggest to continue the flux measurements for a longer
period (e.g. one year) to confirm the representativeness of the fluxes throughout the different
climatic seasons of the year. This would be a valuable contribution for quantifying trace gas
161
emissions that could have regional and global impacts and to reduce the uncertainties due to the
own variability of the emissions. To characterize the three different climatic seasons of Mexico
City, we suggest seasonal measurements of three-months. An analysis of random and systematic
errors using the flux data reported here indicates that a three-months period will yield fluxes
resolved within 11 and 35% of the mean daily flux.
Finally, we recommend that ambient concentrations of individual VOCs need to be measured
in a periodic basis throughout the entire year at different locations of the city with the aim to
determine the seasonal variability and evaluate the effectiveness of the control strategies. Local
authorities should consider the addition to the actual monitoring network of instruments to
measure continuously key hydrocarbons in terms of their reactivity and toxicity, such as toluene
or benzene.
Recommended